JP6338984B2 - Behavior analysis apparatus, behavior analysis method, and program - Google Patents

Description

  Embodiments described herein relate generally to a behavior analysis device, a behavior analysis method, and a program.

  With the widespread use of HEMS (Home Energy Management System), a technique for providing information by estimating behaviors in each household from the power consumption of each household is being studied. For example, a technique has been proposed that provides notification of abnormal conditions, advice for reducing power consumption, and the like based on the power consumption in each home. However, although it is desired to provide information suitable for the lifestyle of the family who is living in a normal life that is not abnormal, it is difficult to appropriately estimate the normal lifestyle.

JP 2010-224800 A JP 2006-106813 A JP-A-9-114900 JP 2010-170419 A JP 2002-304681 A

  The problem to be solved by the present invention is to provide a behavior analysis apparatus, a behavior analysis method, and a program capable of appropriately estimating a lifestyle.

The behavior analysis apparatus according to the embodiment includes a behavior feature extraction unit, a behavior tendency estimation unit, a behavior estimation unit, and a behavior occurrence time zone extraction unit . The behavior feature extraction unit extracts a feature of temporal change of the specific living behavior over a plurality of days based on whether or not the specific living behavior is generated for each time zone in a plurality of days. The behavior tendency estimation unit estimates the tendency of the specific living behavior based on the feature extracted by the behavior feature extraction unit. The behavior estimating unit estimates a living behavior based on information related to the electrical device. The behavior occurrence time zone extraction unit extracts the occurrence time zone of the specific living behavior from the preset analysis target time zones based on the specific life behavior estimated by the behavior estimation unit. The behavior feature extraction unit is characterized by temporal changes in the specific living behavior in the analysis target time zone based on the generation time zones of the specific living behavior in a plurality of days extracted by the behavior occurrence time zone extraction unit. To extract. The behavior estimation unit represents the estimated daily behavior as a behavior label, and outputs behavior label series data in which one or a plurality of behavior labels are arranged in a time series on a daily basis as an estimation result. The behavior occurrence time zone extraction unit refers to the behavior label sequence data, extracts a behavior label representing the specific living behavior from the behavior label sequence data including the behavior label, and in the analysis target time zone. Based on the frequency of the extracted action label and a predetermined threshold, an occurrence time zone of the specific living action in the analysis target time zone is extracted.

1 is a system configuration diagram showing a schematic configuration of a behavior analysis system according to a first embodiment. The block diagram which shows schematic structure of the action analysis apparatus of 1st Embodiment. The figure which shows the measurement data of the electric energy of 1st Embodiment. The figure which shows the data transmitted from the household appliances of 1st Embodiment. The figure which shows the status log | history data of 1st Embodiment. The figure which shows the position of the sensor in the home of 1st Embodiment. The figure which shows the human sensitive sensor measurement data of 1st Embodiment. The figure which shows the action record data for learning of 1st Embodiment. The flowchart which shows the process of the teacher learning at the time of performing the estimation process of the living action of 1st Embodiment. The figure which shows the status log | history data of the household appliances for the cooking use of 1st Embodiment. The figure which shows the example of totalization of the number of hours of the ON state for every household appliances of 1st Embodiment. The figure which shows the example of extraction of the time slot | zone where "cooking" of 1st Embodiment does not generate | occur | produce. It is a figure which shows the image of the discriminator learning of 1st Embodiment. The flowchart which shows the estimation process of the living action of 1st Embodiment. The figure which shows an example of the estimation result of the generation | occurrence | production time zone of the living action of 1st Embodiment. The flowchart which shows the action sequence generation | occurrence | production time slot | zone extraction process of 1st Embodiment. It is a figure explaining totalization of the frequency of the start time of the action label of a 1st embodiment. It is a figure explaining totalization of the frequency of the end time of the action label of a 1st embodiment. The figure which arranged the frequency of the start time of the action label of 1st Embodiment in ascending order. The figure which arranged the frequency of the end time of the action label of a 1st embodiment in ascending order. The figure which shows the time zone division about the frequency of the start time of the action label of 1st Embodiment. The figure which shows the time zone division about the frequency of the end time of the action label of 1st Embodiment. The figure which shows the action generation | occurrence | production time slot | zone area of the specific action label of 1st Embodiment. The figure which shows the example of classification | category of the generation time zone of "cooking" of 1st Embodiment. The figure which shows the example of extraction of the feature-value regarding the action occurrence frequency in the generation | occurrence | production time slot | zone of "cooking" of 1st Embodiment. The figure which shows the action generation | occurrence | production time slot | zone area of the some action label of 1st Embodiment. Explanatory drawing explaining another example which estimates the tendency of the living action of 1st Embodiment. Explanatory drawing of the example from which the generation | occurrence | production time zone of a living action differs with respect to FIG. The figure which shows the registration information of the priority rule of the information delivered of 1st Embodiment. The figure which shows the message provided based on the tendency of the specific living action of 1st Embodiment. FIG. 6 is a tree diagram showing an example of selecting information based on the priority of the first embodiment. The figure which shows the selection rule of the transmission timing further with respect to FIG. The flowchart which shows the process which selects the information and output timing which 1st Embodiment provides. The figure which shows the example of a display of the display part of the information terminal of 1st Embodiment. The block diagram which shows schematic structure of the action analysis apparatus of 2nd Embodiment.

  Hereinafter, a behavior analysis device, an analysis method, and a program according to embodiments will be described with reference to the drawings.

(First embodiment)
Hereinafter, the behavior analysis system according to the first embodiment will be described. FIG. 1 is a system configuration diagram illustrating an example of a schematic configuration of a behavior analysis system 1 in the first embodiment.
The behavior analysis system 1 includes a communication device 11 installed in a home 10 (resident) and a server 30 connected to the communication device 11 via a network 20. In FIG. 1, the communication device 11 of one home 10 is shown, but the server 30 is connected to the communication device 11 installed in each of the plurality of homes 10 via the network 20.

  The network 20 is a network such as the Internet, a LAN (Local Area Network), and a mobile phone network. The communication device 11 is a communication device (for example, HGW (Home GateWay)) that relays an electric product (electric device) in the home 10 and the network 20. The electrical products are, for example, a distribution board 12, a home appliance 13, a sensor 14, an information terminal 15, and the like.

  For example, the communication device 11 relays the main power amount or branch power amount measurement data in the distribution board 12 to the network 20. In addition, the communication device 11 relays measurement data or operation history data of the power consumption of the home appliances 13 such as the electric kettle 13 a, the microwave oven 13 b, and the rice cooker 13 c to the network 20. In addition, the communication device 11 relays measurement data of the sensors 14 such as the human sensor 14 a and the temperature / humidity illuminance sensor 14 b to the network 20. The communication device 11 relays data transmitted from the server 30 to the information terminal 15 via the network 20 and data transmitted from the information terminal 15 to the server 30 via the network 20.

  In addition, although the electric kettle 13a, the microwave oven 13b, and the rice cooker 13c are illustrated in this FIG. 1 as an example of the household appliances 13, these are only shown as examples and are not limited thereto. .

  The human sensor 14a senses whether there is a person in a predetermined detection range, or the movement of a person within the detection range, and outputs measurement data. The temperature / humidity illuminance sensor 14b senses ambient temperature, humidity, and illuminance, and outputs measurement data.

  The information terminal 15 is an example of a terminal device that includes a display unit, and is, for example, a mobile phone, a smartphone, a tablet terminal, a personal computer, a television, or the like. For example, the information terminal 15 displays information transmitted from the server 30 on the display unit. The information terminal 15 may be a dedicated terminal device in the behavior analysis system 1.

  The server 30 includes a communication device 31 and a behavior analysis device 32. The communication device 31 has a communication function for relaying between the behavior analysis device 32 and the network 20. The behavior analysis device 32 exchanges information with the home 10 via the communication device 31 and the network 20.

  For example, the behavior analysis device 32 extracts characteristics of temporal changes in specific living behavior over a plurality of days based on information related to electrical appliances in the home 10. Here, the information related to the electrical product includes, for example, information related to the operation history of the electrical product or information related to the power consumed by the electrical product. In addition, the temporal change of specific living behavior over multiple days is the magnitude of the daily difference in the occurrence time of the living behavior or the daily difference in the length of the occurrence time of the living behavior. . In addition, the behavior analysis device 32 can set an arbitrary analysis time zone in one day, for example, and extract a feature of temporal changes in living behavior. Then, the behavior analysis device 32 estimates the tendency of the living behavior based on the characteristics of the temporal change of the living behavior in the home 10. Moreover, the behavior analysis apparatus 32 provides the information according to the tendency (life style) of a user's life behavior based on the analysis result of this life behavior.

  As shown in FIG. 1, the behavior analysis device 32 may have a part of the configuration of the server 30 or a configuration including all the configurations of the server 30 as equivalent to the server 30.

Hereinafter, the configuration of the behavior analysis device 32 will be described in detail.
FIG. 2 is a block diagram illustrating an example of a schematic configuration of the behavior analysis apparatus 32 according to the first embodiment. The behavior analysis device 32 includes an input unit 100, a storage unit 200, a calculation unit 300, and an output unit 400.

First, the configuration of the input unit 100 will be described.
The input unit 100 includes an electrical product operating state estimation condition input unit 101, an action label estimation condition input unit 102, an action sequence occurrence time zone extraction condition input unit 103, an occurrence time zone feature extraction condition input unit 104, an action A trend estimation condition input unit 105.

  The electrical product operating state estimation condition input unit 101 receives an input of a power threshold value for determining the operating state of each electrical product. This power threshold value determines the operating state of each electrical product based on the measurement data of the power consumption of each electrical product (such as the distribution board 12 and the home appliance 13) acquired by the server 30 from the home 10 via the network 20. It is a threshold for The measurement data of the power consumption amount of each electrical product is determined based on the power threshold value, so that the operating state of each electrical product is represented as a plurality of discrete states. For example, the operating state of each electric product is represented by two states such as ON / OFF. In addition, the operating state of the electrical product may be represented by a plurality of states of 3 or more. For example, the operating state of the rice cooker 13c may be expressed in three states of OFF / heat retention / rice cooking using a plurality of power thresholds.

  The action label estimation condition input unit 102 receives an input of a definition of an action label to be extracted and a condition for generating the action label. The action label represents a daily action such as “sleep”, “going out”, “laundry”, “TV viewing”, “cooking”, and the like. The action label generation condition is a condition for determining whether or not an action label is generated, that is, an estimation condition for estimating a living action.

  For example, in the behavior label estimation condition input unit 102, a threshold for determining whether or not a behavior label is generated is set as a behavior label generation condition. In addition, the behavior label estimation condition input unit 102 sets a negative example sample number with respect to a positive example (for example, random sampling) with respect to learning behavior record data for determination using teacher learning (including semi-teacher learning). The same number as the positive example). In the action label estimation condition input unit 102, the maximum target duration of the action label and the like are set as necessary. Furthermore, in the action label estimation condition input unit 102, a list of home appliances 13 necessary for determining whether or not an action label to be extracted is generated is set. A specific example of the action label generation condition will be described in detail in the description of the calculation unit 300 described later.

  The action sequence occurrence time zone extraction condition input unit 103 accepts an input of an extraction condition when extracting an occurrence time zone of a specific action label from action label series data of a plurality of days. Here, the action label series data is data in which action labels generated on a daily basis are arranged in time series, for example. Specifically, the action sequence occurrence time zone extraction condition input unit 103 includes an extraction condition when the action label occurrence time zone is extracted according to the frequency (number of occurrences) of the action label in each unit time and a predetermined threshold. Set as For example, if the frequency of the action label is equal to or greater than a predetermined threshold, it is assumed that the action label is generated (the action was performed at that time). On the other hand, if the frequency of the action label is less than a predetermined threshold, the action label is not generated (the action is not performed at that time). Based on this extraction condition, the generation time zone of the action label is divided.

  In addition, the action sequence occurrence time zone extraction condition input unit 103 includes, as an analysis target time zone, only on weekdays or holidays, whether or not there is rainfall in a certain time zone (for example, “from 9:00 to just before 18:00”). An extraction condition such as extraction may be set. In addition, the length of the action label series data (for example, “from 8:00 to immediately before 22:00”) and the condition for cutting out the action label series data (for example, “from the start of sleep to immediately before the start of the next sleep”) are extracted. It may be set as a condition.

  The in-occurrence time zone feature extraction condition input unit 104 extracts an extraction condition when extracting features of temporal changes of a specific action label over a plurality of days based on whether or not a specific action label is generated for each time period in a plurality of days. Accepts input. For example, in the occurrence time zone feature extraction condition input unit 104, for the occurrence of a specific action label such as “the ratio of cooking in the morning time zone, the noon time zone, and the evening time zone”. A count condition for counting the ratio is set. Here, the “morning time zone” is set, for example, as “from 7:00 to just before 21:00”. The “daytime zone” is set, for example, as “the evening time zone from 11:00 to 13:00”. “Evening time zone” is set, for example, as “a time zone from 16:00 to 18:00”. Note that the extraction condition for extracting the temporal change feature of a specific action label can be set for each of a plurality of action labels. In addition, in the occurrence time zone feature extraction condition input unit 104, a method for calculating a feature amount of temporal change of a specific action label is set.

  The behavior tendency estimation condition input unit 105 receives an input of an estimation condition for estimating a trend of living behavior. Based on the estimation condition, the tendency of the user's living behavior in the home 10 from the temporal change feature of the specific behavior label extracted by the extraction condition set in the generation time period feature extraction condition input unit 104 Is estimated. Then, service content and recommendation information corresponding to the tendency of the living behavior are distributed to the information terminal 15 in the home 10.

Next, the configuration of the storage unit 200 will be described.
The storage unit 200 includes an electrical product power measurement data storage unit 201, an electrical product operation history data storage unit 202, an electrical product state history data storage unit 203, a human sensor measurement data storage unit 204, and a temperature / humidity illuminance sensor measurement. Data storage unit 205, learning action record data storage unit 206, action label sequence data storage unit 207, action sequence occurrence time zone data storage unit 208, action occurrence time zone feature data storage unit 209, and recommended information storage DB210.

  The electrical product power measurement data storage unit 201 stores measurement data of the amount of power within a certain period of time for each electrical product acquired by the server 30 from the home 10 via the network 20. FIG. 3 is a diagram illustrating an example of power amount measurement data stored in the electrical product power measurement data storage unit 201. This figure shows the measurement data T10 of the electric energy of one day (Weekday-01) on weekdays, and data in a two-dimensional tabular format consisting of rows and columns with each time (0 to 23:00) as a column. It is. For example, in the column of 6 o'clock, measurement data of each electric energy from 6 o'clock to immediately before 7 o'clock is stored. Here, in the first row, measurement data of the main power amount of the distribution board 12 is stored in time series. In the second and subsequent rows, measurement data of the power consumption of each home appliance 13 is stored in time series.

  The electrical product operation history data storage unit 202 stores operation history data of each electrical product in the home 10 within a predetermined time. For example, when the server 30 can acquire operation information of each electrical product from the home 10 via the network 20, operation history data is stored based on the acquired operation information.

  For example, if the electrical product in the home 10 is a device that complies with a communication standard such as “ECHONET Lite”, information obtained by an information collecting device mounted on each electrical product is transmitted from each electrical product. And the server 30 can acquire and accumulate | store the information transmitted from each electric product. The information transmitted from each electric product may be data in which a time stamp at the time of operation is added to the operation content. In addition, when information is transmitted from each electric product at regular intervals, the information transmitted from each electric product may be data in which a time stamp at the time of transmission is added to the operation content.

  FIG. 4 is a diagram illustrating an example of transmission data transmitted from the home appliance 13. The transmission data T20 is an example of operation history data transmitted from the living air conditioner, and includes a time stamp when the operation is performed and an operation content at that time. Here, the power is turned on at 6: 4, the power is turned off at 8:55, the power is turned on at 18:10, and the power is turned off at 22: 5.

  Moreover, transmission data T30 shown in FIG. 4 is an example of operation history data transmitted from the rice cooker, and includes a time stamp when the operation is performed or when the operation state is changed, and an operation content or operation state at that time. It is. Here, the power is turned on at 18:00 to enter the rice cooking mode, the rice cooking mode is switched to the heat-retaining mode at 18:55, and the power is turned off at 22:57.

  The electrical product state history data storage unit 203 stores state history data in which the operation state or operating state of each electrical product per unit time is shown in time series. For example, the electrical product status history data storage unit 203 stores a value obtained by interpolating the operation history data stored in the electrical product operation history data storage unit 202 at regular time intervals (for example, a value immediately before the interpolation target time). The

  Further, the electrical product state history data storage unit 203 may store data on the operating state estimated based on the measurement data of the electric energy stored in the electrical product power measurement data storage unit 201 in time series. For example, data indicating the operating state of each electrical product estimated based on the measurement data of the electric energy and the power threshold value of each electrical product input by the electrical product operating state estimation condition input unit 101 is stored as state history data. Is done.

  That is, the electrical product state history data storage unit 203 stores state history data generated based on one or both of operating state data and operation history data estimated from the power measurement data.

  FIG. 5 is a diagram illustrating an example of status history data stored in the electrical product status history data storage unit 203. The state history data T40 shown in this figure shows the state history data of one day (Weekday-01) on a weekday, and is a two-dimensional table composed of rows and columns with each time (0 to 23:00) as a column. Format data. Each row stores the operation state or operating state of each electrical product for each hour in time series. For example, the operation state or operating state of each electrical product from 6 o'clock to 7 o'clock is stored in the 6 o'clock column.

  Here, in the first row, the operation state or operating state (for example, “ON / OFF”) of the distribution board 12 every hour is stored in time series. In the second and subsequent rows, the operation state or operating state of each home appliance 13 every hour (for example, “ON / OFF”, in the case of a rice cooker, “rice cooking / warming / OFF”) is stored in time series. Has been.

  The human sensor measurement data storage unit 204 stores human sensor measurement data based on the measurement data of the human sensor 14a. FIG. 6 is a diagram illustrating the position of the sensor 14 in the home 10. For example, as shown in this figure, the human sensor 14a is installed in an entrance, a dining & kitchen, a living room, and a private room. In the human sensor measurement data storage unit 204, measurement data sensed in the detection range of each human sensor 14a is stored in time series.

  FIG. 7 is a diagram illustrating an example of human sensor measurement data stored in the human sensor measurement data storage unit 204. The human sensor measurement data T50 shown in this figure indicates time-series measurement data of each human sensor 14a shown in FIG. 6, and is composed of rows and columns with each time (0 to 23:00) as a column. Two-dimensional tabular data. In order from the first line to the fourth line, the measurement data of the human sensors 14a of the entrance, the dining & kitchen (DK), the living room (L), and the private room are stored in time series. Here, as human sensor measurement data, data of time (minutes) when a person is detected in each time (1 hour) is represented.

  In the example shown in FIG. 7, the time (minutes) when the person is detected by the human sensor 14a is stored as human sensor measurement data, but the present invention is not limited to this. For example, data relating to the amount of human movement detected by the human sensor 14a may be stored as human sensor measurement data. The data relating to the amount of movement of the person may be an output value of measurement data output from the human sensor 14a, or data obtained by converting the output value of the human sensor 14a into a movement amount.

  The temperature / humidity illuminance sensor measurement data storage unit 205 stores temperature, humidity, and illuminance measurement data in the home 10 as temperature / humidity illuminance measurement data. For example, as shown in FIG. 6, the temperature / humidity illuminance sensor 14b is installed in a dining & kitchen, a living room, and a private room. In the temperature / humidity illuminance sensor measurement data storage unit 205, temperature / humidity illuminance measurement data of temperature, humidity, and illuminance sensed by each temperature / humidity illuminance sensor 14b is stored in time series.

  For example, when the power consumption amount or ON / OFF state of the lighting device cannot be acquired, whether or not the room is in a bright state that can be used based on the illuminance value measured by the temperature / humidity illuminance sensor 14b. May be determined. Specifically, by providing a threshold value for determining the illuminance value, it is possible to determine whether the room is in a bright state that can be used by illumination by a lighting device or illumination by sunlight from the outdoors. As for humidity, the difference between previous and next humidity values may be calculated, and time series data may be set such that “1” is set when there is a difference greater than a certain threshold, and “0” is set for the other.

  The learning action record data storage unit 206 stores learning action record data generated when the user records which time zone each life action was performed. FIG. 8 is a diagram illustrating an example of learning action record data stored in the learning action record data storage unit 206. The learning action record data T60 shows a record of the user's daily activities on a weekday (Weekday-01), and is a two-dimensional row and column with each time (0 to 23:00) as a column. It is tabular data. In each row, the time periods of occurrence of each action label representing each life action are recorded in time series. Here, “1” is described in the time period of occurrence of each action label. For example, in the third line, the altitude label is “cooking”, and the generation time zone (that is, the time zone during which cooking was performed) is between 6 o'clock and 7 o'clock, between 16:00 and 19 o'clock, Between 20:00 and 21:00 is recorded.

  The learning action record data may be recorded in detail for many action labels as shown in the learning action record data T60. However, as shown in the learning action record data T70, several action labels may be used. It may be recorded only. For example, action labels (here, “sleeping”, “going out”, and “cooking”) in which the user is interested may be recorded, and other action labels may be recorded as action labels such as “others”. Further, the analyst may arbitrarily determine action labels in advance for each time zone of the measurement data.

  The behavior label sequence data storage unit 207 stores behavior label sequence data estimated by the behavior analysis device 32 based on various data acquired from the home 10 by the server 30 via the network 20. Various types of data acquired from the home 10 include power amount measurement data stored in the electrical product state history data storage unit 203 or operation history data stored in the electrical product operation history data storage unit 202. The various data acquired from the home 10 includes human sensor measurement data stored in the human sensor measurement data storage unit 204 or temperature / humidity illuminance measurement data stored in the temperature / humidity illuminance sensor measurement data storage unit 205. May be included. Further, the various data acquired from the home 10 may include learning action record data stored in the learning action record data storage unit 206.

  The estimated action label sequence data is stored in the action label sequence data storage unit 207, and parameter estimation of the estimation model of each action label is also performed using the learning action record data as training data.

  The action label series data stored in the action label series data storage unit 207 is equivalent to the learning action record data stored in the learning action record data storage unit 206.

  In the action sequence occurrence time zone data storage unit 208, among the plurality of action label series data, the action label is considered to occur in a specific time zone but not to occur in a certain other time zone. Data indicating sex is stored.

  The behavior occurrence time zone feature data storage unit 209 stores data indicating characteristics of temporal changes in specific living behavior over a plurality of days. For example, the action occurrence time zone feature data storage unit 209 stores the action occurrence time zones classified based on the occurrence of each specific life activity for each time zone as a temporal change of a specific action label in a plurality of days. Pattern is recorded. The behavior occurrence time zone feature data storage unit 209 includes data indicating the calculation result of the feature amount of the behavior occurrence time zone calculated based on the behavior occurrence frequency of a specific behavior label for each time zone in a plurality of days. Remembered.

  The recommended information storage DB 210 stores information provided according to the user's daily activities. For example, information to be provided includes information such as restaurant coupons and recommended menus, or information on ingredients such as supermarkets. Based on the estimation conditions for estimating the tendency of the living behavior set by the behavior tendency estimation condition input unit 105, it is determined whether or not the information is the home 10 having a certain tendency of the living behavior. It is determined whether or not the information is to be distributed.

  Note that the data stored in each unit of the storage unit 200 may not be acquired directly from the home 10 via the network 20 by the server 30. For example, the server 30 may acquire data transmitted from the home 10 via a data storage server. This data storage server is, for example, a server in which time series data of power sensors connected to each home appliance 13 and time series data of power sensors installed on the distribution board 12 are stored. In addition, the data storage server has time-series data of detection of the moving body by the human sensor 14a, temperature, humidity, or illuminance data, water usage data, gas usage data, etc., measured at regular intervals. It may be a stored server. The data storage server may be divided into a plurality of servers for each type of data stored.

Next, the configuration of the calculation unit 300 will be described.
The arithmetic unit 300 includes an electrical product operating state estimation unit 301, an electrical product state history collection unit 302, an action label estimation unit 303, an action sequence occurrence time zone extraction unit 304, an occurrence time zone feature extraction unit 305, An action sequence occurrence time zone feature extraction unit 306 and an action tendency estimation unit 307 are provided.

  The electrical product operating state estimation unit 301 refers to the power amount measurement data stored in the electrical product power measurement data storage unit 201 and estimates the operating state of each electrical product. For example, the electrical product operation state estimation unit 301 uses the power threshold value input from the electrical product operation state estimation condition input unit 101 to determine the operation of the distribution board 12 or each home appliance 13 from the power amount of each electrical product. Estimate the state. Note that the estimation method may be an estimation method described in, for example, Japanese Unexamined Patent Application Publication No. 2012-254748 or Japanese Unexamined Patent Application Publication No. 2011-39886.

  The electrical product state history collection unit 302 outputs state history data indicating the operation state or operating state of each electrical product in time series. For example, the electrical product state history collection unit 302 obtains the state history data based on the data indicating the estimated operating state of each electrical product and the operation history data stored in the electrical product operation history data storage unit 202. Generate. Then, the electrical product state history collection unit 302 outputs and stores the generated state history data in the electrical product state history data storage unit 203.

  As a result, in the electrical product state history data storage unit 203, action labels are discretely attached to each electrical product every unit time (for example, every hour), for example, as in the state history data T40 shown in FIG. The obtained time series data is stored.

  The behavior label estimation unit 303 estimates the living behavior of the user in the home 10 based on information related to the electrical product in the home 10. Furthermore, the behavior label estimation unit 303 may further estimate the living behavior based on information indicating the presence or absence of a person at a place where the living behavior is performed, or information indicating the temperature, humidity, or illuminance of the place. Thus, the estimation accuracy of living behavior can be improved by adding information indicating the presence or absence of a person or information indicating the temperature, humidity, or illuminance of the place.

  Specifically, the behavior label estimation unit 303 refers to the electrical product state history data storage unit 203, the human sensor measurement data storage unit 204, and the temperature / humidity illuminance sensor measurement data storage unit 205, and performs the living behavior in the home 10. As several discrete action labels. For example, the behavior label estimation unit 303 represents the estimated daily behavior as a behavior label, and outputs behavior label sequence data in which one or a plurality of behavior labels are discretely arranged in time series on a daily basis as an estimation result. The action label estimation unit 303 stores the output action label series data in the action label series data storage unit 207. The action label series data stored in the action label series data storage unit 207 is, for example, data having the same configuration as the learning action record data T60 and T70 shown in FIG.

  As described above, the action label representing the living action includes “sleep”, “going out”, “laundry”, “TV viewing”, “cooking”, and the like. These action labels and the generation conditions of the action labels are input to the action label estimation condition input unit 102 and set. For example, the living behavior is estimated based on the following behavior label generation conditions.

  “Sleep”: If the home appliance 13 other than the air conditioner is in an OFF state and the measurement data of the human sensor 14a has only an output that can be detected when turning over, it is estimated as “sleep”. The

  “Outing”: When all the home appliances 13 are in an OFF state and the measurement data (output value) of the human sensor 14a is finally lowered at the entrance, it is estimated as “outing”. The case where the measurement data (output value) of the human sensor 14a has finally become lower at the entrance means that a person has been detected at the entrance and no longer has been detected in the home 10. In this case, “going out” is assumed.

“Laundry”: “Washing” is estimated when the washing machine is ON.
“TV watching”: When the TV is ON, it is estimated that “TV watching”.

“Cooking”: Since a plurality of home appliances 13 are used and the temporal relationship of the operating states of the plurality of home appliances 13 is related, the following estimation method is used.
First, the action label estimation condition input unit 102 sets a list of home appliances 13 used for cooking purposes as a list of home appliances 13 necessary for determining whether or not “cooking” has occurred. For example, this list is set when an analyst inputs the action label estimation condition input unit 102 in advance. In addition, the action label estimation unit 303 sets the time series time interval between the state history data of the home appliance 13 used for cooking and the learning action record data, and records that the user actually cooks. Data.

  The lifestyle activity estimation process performed by the behavior label estimation unit 303 will be described with reference to FIGS. 9 to 14. FIG. 9 is a flowchart illustrating an example of a teacher learning process when performing a daily activity estimation process.

  First, the action label estimation unit 303 extracts the occurrence time zone of “cooking” from the action history data for learning from the state history data. Here, the state history data is, for example, the state history data T40 stored in the electrical product state history data storage unit 203. The learning action record data is, for example, learning action record data T60 stored in the learning action record data storage unit 206. For example, the action label estimation unit 303 extracts the occurrence time zone of “cooking” from the state history data T40 in the learning action record data T60 (in the example of FIG. 8, the time zone in which “1” is described) ( Step S101).

  Next, among the time periods in which “cooking” occurs, the time period in which there is no ON-use household appliance 13 for cooking is maintained immediately before the start of “cooking” or from the end of “cooking” Are excluded from the teacher data (step S102).

  Here, the processing of step S101 and step S102 will be described with reference to FIG. The state history data T80 shown in FIG. 10 is state history data of the household appliance 13 for cooking use in the state history data T40 shown in FIG. Further, in the state history data T90, in the state history data T80, a column of 6 o'clock and a column of 16 o'clock to 18 o'clock are shown as an example of the occurrence time zone of “cooking”. The 19 o'clock column is a period in which there is no cooking appliances 13 in the cooking state at the end of “cooking” and after that. In other words, the 19 o'clock column is a period in which the time zone in which there is no cooking-use household appliance 13 in the ON state has continued from the end of “cooking”. Therefore, this 19 o'clock column period is excluded from the teacher data. And it forms as an estimation model for every time continuing as "cooking" (F1, F2, F3, ...).

  Next, the action label estimation unit 303 selects a discriminator according to the length of the “cooking” occurrence time zone (step S103). Then, the behavior label estimation unit 303 counts the number of hours (number of columns) in the ON state for each household electrical appliance 13 for cooking in the “cooking” occurrence time zone (step S104).

  FIG. 11 is a diagram illustrating an example of counting the number of hours in the ON state for each home appliance 13. This figure shows an example of counting in the case of a discriminator whose generation time zone is 3 hours long. In the example shown in this figure, in the row from 16:00 to 18:00, the total result of the number of hours in the ON state of each of the electric kettle 13a, the microwave oven 13b, the rice cooker 13c, and the toaster is “0, 2, 1, Represented as "0".

  Next, the action label estimation unit 303 extracts a time zone in which “cooking” has not occurred in the learning action data (step S105). The behavior label estimation unit 303 sequentially extracts the state history data of the time period in which “cooking” does not occur by sliding one column at a time for each length of the time period in which “cooking” occurs (step S106).

  FIG. 12 is a diagram illustrating an extraction example of a time period when “cooking” does not occur. The example shown in FIG. 12 is an extraction example in the case of a discriminator whose generation time zone is 3 hours long. The time period when the “cooking” of the state history data T90 is not generated is slid by one row every three hours, and the number of hours in which each of the electric kettle 13a, the microwave oven 13b, the rice cooker 13c, and the toaster is ON is determined. Aggregated. For example, the totaling result when none is in the ON state as in the column from 0 o'clock to 2 o'clock, for example, is represented as “0, 0, 0, 0”. Further, for example, as in the column from 20:00 to 22:00, the total result when only the electric kettle 13a is ON for one hour is represented as “1, 0, 0, 0”. In this example, the tally results for the time period when “cooking” does not occur are twelve tally results represented as “0, 0, 0, 0” and “1, 0, 0, 0”. And one totaling result is obtained.

  Next, the action label estimation unit 303 selects some data as learning data from the state history data in a time zone in which “cooking” does not occur (step S107). For example, in the example illustrated in FIG. 12, nine of 12 pieces of state history data whose total result is “0, 0, 0, 0” and state history data which is “1, 0, 0, 0”. Is selected as learning data.

  Then, the action label estimation unit 303 inputs the state history data in the time period of occurrence of “cooking” as a positive example and the state history data of a time period in which some “cooking” does not occur as a negative example to the discriminator. To learn (step S108). FIG. 13 is a diagram illustrating an image of discriminator learning in which a positive example and a negative example are input to the discriminator to learn.

  In this way, the action label estimation unit 303 identifies based on the state history data of the electrical appliance (excluding the time zone in which there is no device ON state at both ends of the teacher data) and the positive and negative examples of the teacher data. The vessel is constructed by teacher learning. And the action label estimation part 303 estimates the lifestyle action of cooking from the state log | history data of the household appliances 13 for cooking use using the presumed conditions learned by the discriminator.

  Next, with reference to FIG. 14, an estimation process for estimating a living behavior using a discriminator based on teacher learning will be described. FIG. 14 is a flowchart showing a daily behavior estimation process based on teacher learning.

  The action label estimation unit 303 sweeps the state history data of the household appliances 13 for cooking and searches for a time zone when any one of the household appliances 13 is in an ON state (step S201). Next, the action label estimation unit 303 uses the discriminator according to the length of the “cooking” time zone learned by the teacher learning, as the data corresponding to “cooking”. Is determined (step S202).

  And the action label estimation part 303 estimates the time slot | zone corresponding to "cooking" as an occurrence time slot | zone of "cooking". Moreover, when the time slot | zone corresponding to "cooking" continues, the action label estimation part 303 determines the continuous longest time slot | zone as one "cooking" generation time slot | zone (step S203).

  FIG. 15 is a diagram illustrating an example of an estimation result of the occurrence time zone of the living behavior. In the example shown in this figure, the 6 o'clock column and the 17 o'clock to 18 o'clock column are estimated from the state history data as “cooking” occurrence time zones.

  Thus, for example, the behavior label estimation unit 303, based on the state history data and the living behavior recording data in which the actual living behavior for each time zone is recorded, the estimation condition of the living behavior across a plurality of time zones. Teacher learning about. And the action label estimation part 303 estimates the living action over a several time slot | zone based on the estimation conditions which performed teacher learning.

  By using teacher learning in this way, normal living behavior in the home 10 can be accurately estimated. In particular, it is possible to improve the estimation accuracy of living behavior in which a plurality of home appliances 13 are used like “cooking”.

  In the above-described example, the example in which the action label estimation unit 303 outputs action label series data for each day as an estimation result has been mainly described. Thereby, although the behavior analysis apparatus 32 can compare the generation | occurrence | production time slot | zone of the living action in multiple days per day, it is not restricted to this. For example, the behavior label estimation unit 303 may output the behavior label series data with one unit from the beginning of the behavior label indicating “sleep” to immediately before the next behavior label indicating “sleep” as an estimation result. . Thereby, the behavior analysis device 32 can cope with various lifestyles such as a user who is active at night and sleeps in the daytime, a user who has night shift and day shift depending on the day.

  Next, the action sequence occurrence time zone extraction unit 304 will be described. The action sequence occurrence time zone extraction unit 304 extracts the occurrence time zone of the specific living behavior from the preset analysis target time zones based on the specific life behavior estimated by the behavior label estimation unit 303. The specific living behavior is, for example, a living behavior selected as an object to be analyzed by the analyst.

  Specifically, the action sequence occurrence time zone extraction unit 304 refers to the action label series data and extracts an action label representing a specific living action from action label series data including the action label. Moreover, the action sequence occurrence time zone extraction unit 304 extracts the occurrence time zone of a specific living action in the analysis target time zone based on the frequency of the action label extracted in the analysis target time zone and a predetermined threshold. Here, the predetermined threshold is an action occurrence area determination threshold described later.

  Hereinafter, the operation of the action sequence occurrence time zone extraction process for extracting the occurrence time zone of the specific living behavior will be specifically described with reference to FIGS. 16 to 22. FIG. 16 is a flowchart illustrating an example of an action sequence occurrence time zone extraction process according to the present embodiment.

  First, the action sequence occurrence time zone extraction unit 304 receives the action label sequence data of one type and a plurality of days, and extracts the frequency (number of times) of the start time or end time of the action label for each time zone. For example, the behavior sequence occurrence time zone extraction unit 304 refers to the behavior label sequence data storage unit 207 and collects data of a specific behavior label to be analyzed and analysis target time zone from the behavior label sequence data. Here, the specific action label and analysis target time zone to be analyzed are those set by the action sequence occurrence time zone extraction condition input unit 103 (step S301).

  Next, the action sequence occurrence time zone extraction unit 304 aggregates the start (or end) time of a specific action label as a frequency for each time from the collected data (step S302).

FIG. 17 and FIG. 18 are diagrams for explaining the aggregation of the frequency of the start time and end time of the action label for each time zone.
FIG. 17 shows an aggregation of the start times of certain action labels in a certain specific time zone. In the graph indicated by symbol a, the horizontal axis represents time, the vertical time represents the action occurrence ID, and represents the occurrence time zone of the action label for each action occurrence ID. The action occurrence ID is an identifier that can identify an action label for each day, for example. In the graph indicated by reference sign b, the horizontal axis is time, the vertical time is the frequency of the start time of the action label, and the start time of the action label indicated by reference sign a is tabulated as the frequency for each time.

  FIG. 18 shows an aggregation of end times of a certain action label. In the graph indicated by symbol a, the horizontal axis represents time, the vertical time represents the action occurrence ID, and represents the occurrence time zone of the action label for each action occurrence ID. In the graph shown by the symbol b, the horizontal axis is the time, the vertical time is the frequency of the action label hunting time, and the action label start time indicated by the symbol a is tabulated as the frequency for each time.

  For example, the action sequence occurrence time zone extraction unit 304 uses the method described above, and the frequency of the start time of the action label of a specific living action for the analysis target time zone set in advance by the action sequence occurrence time zone extraction condition input unit 103. And the frequency of the end time.

  In addition, the action sequence occurrence time zone extraction unit 304 sorts the tabulated data for each time in descending order of frequency (step S303). For example, the action sequence occurrence time zone extraction unit 304 arranges the frequency of the start time and the frequency of the end time of the above-described action labels in ascending order. FIG. 19 is a diagram in which the frequency of the start time of the action label is arranged in ascending order. FIG. 20 is a diagram in which the frequencies of the end times of the action labels are arranged in ascending order.

  Next, in steps S304 to S308, the action sequence occurrence time zone extraction unit 304 performs the processing of steps S305 to S307 for the total number of times sorted in step S303. Then, when the total number is completed, the action sequence occurrence time zone extraction unit 304 proceeds to the process of step S309.

  First, in step S305, the action sequence occurrence time zone extraction unit 304 adds the frequency (the highest frequency) of the selected times among the sorted times to the “number of action occurrences”. Next, in step S306, the action sequence occurrence time zone extraction unit 304 determines whether or not “behavior occurrence number / total action occurrence number + action non-occurrence ratio threshold> 1” is satisfied. If it is determined in step S306 that it is not satisfied (NO), the action sequence occurrence time zone extraction unit 304 returns to step S305, selects the next highest frequency from the sorted times, and adds it to the action occurrence number. To do. On the other hand, when it is determined in step S306 that the event has been satisfied (YES), the action sequence occurrence time zone extraction unit 304 determines the frequency of the time added at this time as an action occurrence area determination threshold (see FIG. 19 or FIG. 20). . Then, the action sequence occurrence time zone extraction unit 304 stores the determined action occurrence region determination threshold in the action sequence occurrence time zone data storage unit 208 (step S307).

  In this way, the action sequence occurrence time zone extraction unit 304 uses the behavior non-occurrence rate threshold value set by the action sequence occurrence time zone extraction condition input unit 103, and determines the occupancy rate of the occurrence rate from the one with the highest frequency. It is determined whether or not the sum of the behavior non-occurrence ratio threshold exceeds 1. Then, the action sequence occurrence time zone extraction unit 304 sequentially adds the next highest frequency when it does not exceed 1, and determines the frequency added when the value exceeds 1 as the action occurrence region determination threshold. FIG. 19 shows the action occurrence region determination threshold value determined by the frequency of the start time of the action label. FIG. 20 shows a behavior occurrence region determination threshold value determined by the frequency of the action label end time.

  Next, in step S309, the action sequence occurrence time zone extraction unit 304 determines whether an action occurrence area determination threshold is stored. If it is determined in step S309 that the action occurrence region determination threshold is not stored (NO), the action sequence occurrence time zone extraction unit 304 ends the action sequence occurrence time zone extraction process. On the other hand, when it is determined in step S309 that the action occurrence region determination threshold is not stored (YES), the action sequence occurrence time zone extraction unit 304 proceeds to the process of step S310.

  In steps S310 to S314, the action sequence occurrence time zone extraction unit 304 performs the processes of steps S311 to S313 for the total number of times of a specific action label.

  First, in step S311, the action sequence occurrence time zone extraction unit 304 sequentially determines whether the frequency of the start (or end) time of a specific action label is equal to or greater than an action occurrence area determination threshold. If it is determined in step S311 that it is equal to or greater than the action occurrence region determination threshold (YES), the action sequence occurrence time zone extraction unit 304 registers the determined time as the action occurrence time zone (step S312). On the other hand, if it is determined in step S311 that it is less than the action occurrence region determination threshold (NO), the action sequence occurrence time zone extraction unit 304 registers the determined time as an action non-occurrence time zone (step S313). .

  Then, when all the times of specific action labels have been completed, the action sequence occurrence time zone extraction unit 304 proceeds to the process of step S315 (step S314).

  In step S315, the action sequence occurrence time zone extraction unit 304 divides the time zone into two adjacent times between a time that exceeds the action occurrence region determination threshold and a time that does not exceed the action occurrence region determination threshold. Then, the action sequence occurrence time zone extraction unit 304 assigns an action occurrence time zone label for a time that exceeds the action occurrence area determination threshold in each of the divided sections. Then, the action sequence occurrence time zone extraction unit 304 stores the start time zone time (or end time zone time) and the action occurrence time zone label in each section of the specific action label in the action sequence occurrence time zone data storage unit 208. (Step S316).

  In this way, the action sequence occurrence time zone extraction unit 304 uses the determined action occurrence area determination threshold value to determine whether or not the frequency at each start time and the frequency at each end time exceed the action occurrence area determination threshold value. Do. Then, the action sequence occurrence time zone extraction unit 304 divides the time zone into sections between the time when the threshold is exceeded and the time when the threshold is not exceeded. Hereinafter, the time zone of each zone divided into zones is also referred to as “time zone classification”.

  FIG. 21 is a diagram illustrating time zone divisions regarding the frequency of the start time of the action label. As shown in this figure, the time when the action occurrence region determination threshold is not exceeded is regarded as the action label is not generated, and is divided between the time when the threshold is exceeded and the time when the action label is not exceeded. Here, the time zone is divided into “Ps0”, “Ps1”, “Ps2”, and “Ps3 (= Ps0)”. “Ps0” is a time zone from 0 o'clock to 7 o'clock, and “Ps1” is a time zone from 7 o'clock to 11 o'clock. “Ps2” is a time zone from 11:00 to immediately before 17:00, and “Ps3 (= Ps0)” is a time zone from 17:00 to immediately before 24:00 (0 o'clock).

  FIG. 22 is a diagram illustrating time zone divisions regarding the frequency of the end time of the action label. As shown in this figure, the time when the action occurrence region determination threshold is not exceeded is regarded as the action label is not generated, and is divided between the time exceeding the threshold and the time not exceeding. Here, the time zone is divided into “Pe4”, “Pe3”, “Pe2”, and “Pe1”. “Pe4” is a time zone from 0:00 to just before 9:00, and “Pe3” is a time zone from 9:00 to just before 14:00. “Pe2” is a time zone from 14:00 to 20:00, and “Pe1” is a time zone from 20:00 to 24:00 (0 o'clock).

  In this way, the behavior analysis device 32 extracts the occurrence time zone of the living behavior based on the frequency of the behavior label and the behavior occurrence region determination threshold. Thereby, since the behavior analysis apparatus 32 can exclude the living behavior with a low occurrence frequency from the feature, it is possible to improve the estimation accuracy of the tendency of the living behavior described later.

  Next, the occurrence time zone feature extraction unit 305 will be described. The feature extraction part 305 between generation time zones extracts the feature of the temporal change of the specific living behavior over a plurality of days based on the presence / absence of the specific living behavior for each time zone in a plurality of days. For example, the inter-occurrence time period feature extraction unit 305 displays the characteristics of temporal changes in specific living behavior in the analysis target time period based on the action sequence generation time period extracted by the action sequence occurrence time period extraction unit 304. Extract.

  Specifically, the inter-occurrence time zone feature extraction unit 305 classifies each of the specific living behaviors in a plurality of days for each combination in the same occurrence time zone based on the occurrence time zone of the specific living behavior. Then, the occurrence time zone feature extraction unit 305 extracts a feature of temporal change in specific living behavior in the analysis target time zone based on the classified result. In this case, the characteristic of the temporal change of the specific living behavior is, for example, the pattern of the occurrence time zone classified based on the occurrence / non-occurrence of the specific living behavior for each time zone.

  Specifically, the feature extraction unit 305 between occurrence times is based on the action sequence occurrence time zone extracted by the action sequence occurrence time zone extraction unit 304. The label series pattern is classified, and each action occurrence time zone section is set. Thereby, the behavior analysis device 32 can classify and represent the characteristics of the temporal change of the living behavior by the combination of the generation time zones.

  FIG. 23 is a diagram illustrating an example of an action occurrence time zone section of a specific action label. In this figure, an action start occurrence time zone and an action end occurrence time zone of a specific action label (for example, “cooking”) in a certain day (from 0:00 to 23:00) are shown. Each of “Ps0”, “Ps1”, “Ps2”, and “Ps3” indicates the time zone division of the start time as shown in FIG. Each of “Pe4”, “Pe3”, “Pe2”, and “Pe1” indicates the time zone classification of the end time as shown in FIG. Moreover, in this figure, each action generation | occurrence | production label in case the data of "cooking" for 12 days (C-01-C-12) is acquired is shown in time series.

  The occurrence time zone of each of the 12 days of “cooking” (C-01 to C-12) in one day is classified into six ways when classified by the combination of the time zone division of the start time and the time zone division of the end time. being classified. FIG. 24 is a diagram illustrating a classification example of the occurrence time zone of “cooking”. In the example shown in this figure, the time zone division between the start time and the end time is (Ps1, Pe4), (Ps0, C-Pe2), (Ps2, Pe3), (Ps2, Pe2), (Ps0, Pe1). ) Are classified into six types. That is, it can be seen that the time zone in which “cooking” occurs is the above-described six time zones. The occurrence time zone feature extraction unit 305 causes the behavior occurrence time zone feature data storage unit 209 to store the result of classifying the time zone classification of these action occurrence time zones.

  The action sequence occurrence time zone feature extraction unit 306 is based on the generation time zones of the specific living activities in a plurality of days extracted by the action sequence occurrence time zone extraction unit 304, and the time of the specific living behavior in the analysis target time zone The characteristics of typical changes. For example, the behavior sequence occurrence time zone feature extraction unit 306 calculates a feature amount related to the behavior occurrence frequency in a certain time zone as a feature of temporal change of a specific living behavior over a plurality of days.

  Specifically, the action sequence occurrence time zone feature extraction unit 306 includes a start time, an end time, an average and variance of durations, and a high frequency time zone within each occurrence time zone of a specific living activity on a plurality of days. One or more of the frequency occupancy rates are extracted. Thereby, the behavior analysis apparatus 32 can represent the feature of the temporal change of the living behavior as a feature value obtained by quantification.

  FIG. 25 is a diagram illustrating an example of extracting feature amounts related to the behavior occurrence frequency in the “cooking” occurrence time zone. This figure is an example of extracting feature amounts related to the action occurrence frequency of “cooking” shown in FIG. Here, the start time, the end time, the duration, the average, the variance, and the like are extracted as feature amounts for each time zone segment, regardless of the time zone segment classification shown in FIG.

Note that an average, variance, and the like may be calculated for each pattern classified according to the time zone classification and extracted as a feature amount. In addition, analysis days are selected for each state of factors that influence daily activities such as weekdays / holidays, whether there is rainfall, whether there are events, etc., and the average, variance, etc. are calculated for each time zone of the analysis day Then, it may be extracted as a feature amount. These various conditions are set in advance as extraction conditions in the occurrence time zone feature extraction condition input unit 104.
The action sequence occurrence time zone feature extraction unit 306 stores these extracted feature amounts in the action occurrence time zone feature data storage unit 209.

  The behavior tendency estimation unit 307 estimates a tendency of specific living behavior based on the features extracted by the occurrence time zone feature extraction unit 305 or the behavior sequence occurrence time zone feature extraction unit 306. For example, the behavior tendency estimation unit 307 refers to the behavior occurrence time zone feature data storage unit 209 and the recommended information accumulation DB 210 to determine what tendency of the lifestyle behavior exists from the current lifestyle behavior characteristics of the home 10. Label with the behavior trend label shown. Then, the behavior tendency estimating unit 307 selects information suitable for the labeling result.

A specific example of the estimation method will be described below.
FIG. 26 is a diagram illustrating an example of an action occurrence time zone section of a plurality of action labels. FIG. 23 shows an example of one specific action label (for example, “cooking”), but FIG. 26 shows the occurrence time zones of “sleep”, “going out”, and “cooking”. Each of “S-Psα” and “S-Peα” indicates a time zone division of the start time and a time zone division of the end time of “sleep” (“α” is 0 or more, respectively) integer). Further, “A-Psα” and “A-Peα” respectively indicate the time zone division of the start time of “outing” and the time zone division of the end time. Further, “C-Psα” and “C-Peα” respectively indicate the time zone division of the start time and the time zone division of the end time of “cooking”.

The behavior tendency estimation unit 307 may estimate the tendency of living behavior based on the occurrence frequency (frequency magnitude) for each time zone. For example, when the following “condition 1” is satisfied, the behavior tendency estimation unit 307 labels “cooking group at home” as the behavior tendency label.
“Condition 1”: The ratio of the number of days in which “cooking” living behavior occurs once or more per day is 0.8 or more.

  In addition, the behavior tendency estimation unit 307 may estimate the tendency of living behavior based on the feature amount extracted by the behavior series occurrence time zone feature extraction unit 306 (for example, “condition 2”, “condition 3” below) ”And“ Condition 4 ”).

For example, when the following “condition 2” is satisfied, the behavior tendency estimation unit 307 labels “dinner cooking group at home” as the behavior tendency label.
“Condition 2”: The ratio of the number of days in which “cooking” living behavior occurs at least once in the day from 15:00 to 21:00 is 0.6 or more.

Further, the behavior tendency estimation unit 307 labels “going out in the evening” as the behavior tendency label when the following “condition 3” is satisfied.
“Condition 3”: The ratio of the number of days in which a “going out” living action occurs once or more from 15:00 to 21:00 in one day is 0.5 or more.

Further, the behavior tendency estimation unit 307 labels “dinner cooking time tight group” as the behavior tendency label when the following “condition 4” is satisfied.
“Condition 4”: All combinations of time series divisions (C-Psα, C-Peα) on the day when the daily activity of “cooking” occurs more than once in the day from 15:00 to 21:00 The appearance ratio of the top time of the “cooking” end time is 0.8 or more of (C-Psα, C-Peα).

In addition, the behavior tendency estimation unit 307 may estimate the tendency of living behavior based on the relationship between the generation time zones of a plurality of behavior labels. For example, when the following “condition 5” is satisfied, the behavior tendency estimation unit 307 labels “cooking to sleep approach group” as the behavior tendency label.
“Condition 5”: The day when the “cooking” living action occurs at least once from 15:00 to 21:00 in the day, and the time of the last “cooking” end time and “sleep” in the day The ratio of intervals less than 1 hour is 0.3 or more.

  The behavior tendency estimation unit 307 may set the timing for transmitting information based on the occurrence frequency (frequency magnitude) for each time zone. For example, the behavior tendency estimation unit 307 provides information corresponding to the tendency of the lifestyle behavior to the home 10 during the time period when the occurrence rate of the lifestyle behavior of “sleep”, “going out”, and “cooking” is less than 0.2. May be selected as the transmission timing. In the example shown in FIG. 26, the time period in which the occurrence rate of “sleeping”, “going out”, and “cooking” is less than 0.2 is “from 8:00 to immediately before 12:00”, “from 14:00” Time zones from “until just before 16:00” and “from 20:00 to just before 22:00”.

  FIG. 27 is an explanatory diagram for explaining another example of estimating a tendency of living behavior. This figure shows only data relating to “cooking” in FIG. In this figure, the “cooking” tendency is estimated from the living behavior of “cooking” in which cooking is completed from 17:00 to 21:00 (from 17:00 to immediately before 22:00).

  For example, out of the data of “cooking” (C-01 to C-12) for 12 days (12 times), the occurrence of “cooking” in the time zone division (C-Ps0, C-Pe2) is 10 times. There is no occurrence twice. From this, the behavior tendency estimation unit 307 labels “there are many cooking opportunities”, “there are few cooking patterns”, “the cooking time zone is stable”, and the like as the behavior tendency labels. Further, the behavior tendency estimation unit 307 may label “cooking time adjustment consciousness is high” because the dispersion of the end time is small. Further, the behavior tendency estimation unit 307 may label “there are many cooking repertoires” as the behavior tendency label since the variance of the time taken for cooking is large.

  FIG. 28 is an explanatory diagram of an example in which the occurrence time zone of the living behavior is different from that in FIG. In this figure, similar to the example described with reference to FIG. 27, from the lifestyle action of “cooking” where cooking is completed at 17:00 to 21:00 (from 17:00 to immediately before 22:00), “cooking” A trend is estimated.

  In this example, the occurrence of “cooking” in the time zone division (C-Ps2, C-Pe2) is 1 in the data of “cooking” (C-01 to C-12) for 12 days (12 times). The occurrence of “cooking” in the time zone division (C-Ps3, C-Pe2) is twice. In addition, the occurrence of “cooking” in the time zone division (C-Ps0, C-Pe2) is once and is not generated twice. As a result, the behavior tendency estimation unit 307 labels, for example, “Many cooking opportunities”, “Many cooking patterns”, and “Cooking time zone is unstable” as behavior tendency labels.

  Also, the behavior tendency estimation unit 307 may label “cooking time adjustment consciousness is low” because the dispersion of end times is large. Also, the behavior tendency estimation unit 307 may label “Many cooking repertoires” because the variance of the time taken for cooking is large.

  Then, the behavior tendency estimation unit 307 uses the combination of behavior tendency labels described with reference to FIG. 27 or FIG. 28, and a priority rule that prescribes which information is preferentially distributed is stored in advance as recommended information. It may be registered in the DB 210.

  FIG. 29 is a diagram illustrating an example of registration information of priority rules for information to be distributed. Here, for example, there is recommended information such as “restaurant restaurant introduction”, “meal food introduction”, “food special sale”, “introduction of maintenance products such as storage containers”, “cooking advice”, etc. . Which of the recommended information is preferentially distributed is defined by a combination of action tendency labels.

  Here, as the behavior tendency label, “cooking opportunities” are many (○) or few (×), “cooking time stability” is high (○) or low (×), or “time adjustment awareness” is high ( It is defined by a combination of (○) or low (×), or “repertoire” having many (○) or few (×).

  For example, it is assumed that there are many “cooking opportunities” (◯), “cooking time stability” is high (◯), “time adjustment awareness” is high (○), and “repertoire” is many (○). In this case, the priority of “restaurant introduction”, “meal meal introduction”, and “cooking method advice” is low. On the other hand, the priority of “sale of special ingredients” and “introduction of maintenance products such as storage containers” are high.

  As described above, the behavior tendency estimation unit 307 refers to the recommended information accumulation DB 210 in which a plurality of candidates for information to be provided to a user related to a specific living behavior are stored. And the action tendency estimation part 307 determines the priority of the information provided to the user relevant to specific living action among the said several candidates based on the tendency of the estimated specific living action. Thereby, the behavior analysis apparatus 32 can preferentially provide information suitable for the lifestyle.

Next, the configuration of the output unit 400 will be described.
The output unit 400 includes an action tendency reflection recommendation information output unit 401. The behavior tendency reflection recommendation information output unit 401 outputs information to be provided to the user related to the specific life behavior based on the tendency of the specific life behavior estimated by the behavior trend estimation unit 307.

  Further, the behavior tendency reflection recommendation information output unit 401 may output information provided to the user related to the specific living behavior at a time designated based on the generation time zone of the specific living behavior. For example, the behavior tendency reflection recommendation information output unit 401 may output information to be provided at a specified time in the generation time zone of the living behavior. Further, the behavior tendency reflection recommendation information output unit 401 may output information to be provided outside the designated time in the generation time zone of the living behavior. Thereby, the behavior analysis apparatus 32 can provide information suitable for life at an appropriate timing.

Hereinafter, a selection example of information to be provided based on a tendency of specific living behavior will be specifically described.
FIG. 30 is a diagram illustrating an example of a message provided based on a tendency of specific living behavior. In the example shown in this figure, a home feature ID, an action tendency label, a message, and an adoption rate are associated with each other. The home characteristic ID is identification information given to each definition of the characteristic of the tendency of the living behavior. Here, the behavior tendency label is an example of a combination of “cooking opportunity”, “cooking time stability”, “time adjustment awareness”, and “repertoire” as in the example of FIG. The message is a candidate message to be provided. The adoption rate is a probability of being adopted as a message actually provided from each message candidate in the same home characteristic ID. These pieces of information are stored in advance in the recommended information storage DB 210.

  For example, the home characteristic ID “01” has many “cooking opportunities” (◯), “cooking time stability” is high (○), “time adjustment awareness” is high (○), and “repertoire” is many (○ ) A candidate message to be provided to the home 10 having a combination feature is defined. Here, “Thank you for your feast!” And “Conventional cooking!” Are defined as message candidates. The former adoption rate is “0.8” and the latter adoption rate is “ 0.2 ".

  Also, the behavior tendency reflection recommendation information output unit 401 selects and outputs information to be actually provided based on the priority of information to be provided determined by the behavior tendency estimation unit 307 (see, for example, FIG. 29).

  FIG. 31 is a tree diagram showing an example of selecting information based on the priority of information to be provided. In this figure, the home feature ID is a route, each home feature ID has a recommended genre as a lower node, and each recommended genre has advertisement content as a lower node. The selection probability of which recommended genre is selected from each home feature ID is set in advance based on, for example, the priority of information provided by the behavior tendency estimation unit 307 shown in FIG.

  For example, the selection probabilities of selecting each recommended genre (“restaurant” (restaurant introduction), “food sales”, “meal” (meal introduction))) from the home feature ID “01” are “0.35”, respectively. , “0.35”, “0.30”. In addition, when “restaurant” is selected as the recommended genre, advertisement content candidates to be further selected are set. For example, as an advertisement content candidate, “Sometimes slowly outside, × □ Daikanyama” (recruitment rate: 0.3), “Authentic Spanish taste, Grazzie ☆ □” (recruitment rate: 0.15), ...・ Etc. are set.

  That is, the home feature ID that matches the combination of behavior tendency labels estimated for the home 10 to be analyzed is, for example, the home feature ID “01”. In this case, as a recommended genre, one of “restaurant”, “food special sale”, and “garden” is selected with selection probabilities of “0.35”, “0.35”, and “0.30”, respectively. For example, when “restaurant” is selected as the recommended genre, the advertisement content to be actually provided is selected from the advertisement content candidates according to the above-described adoption rate.

  FIG. 32 is a diagram showing a rule for selecting a transmission timing for the advertising content whose recommended genre is “dining out” in FIG. 31. Like the selection rule shown in this figure, the behavior analysis device 32 can set the transmission timing of the advertising content for each advertising content candidate based on the information on the time of occurrence of each behavior label. Here, for each advertisement content, a candidate for occurrence timing and its selection probability are set as selection rules. For example, in the case of the advertisement content “Sometimes slowly outside, X □ Daikanyama”, two types of transmission timing candidates are set, and the advertisement content is sent from the behavior analysis device 32 at the transmission timing selected according to each selection probability. Provided to home 10. Here, the period from the end of cooking behavior that occurs mainly in the morning to 12:00 (selection probability: 0.4) and the period from the occurrence of “going out” to the occurrence of “sleep” (Selection probability: 0.6) is set as a candidate for the transmission timing. The period from the end of the cooking action that occurs mainly in the morning to 12:00 is, for example, the period from the end of the “cooking” end time zone that occurred from 0 to 10:00 until 12:00. , Is set as In addition, the period from the occurrence of “going out” to the time before the occurrence of “sleep” is, for example, after the end of the end time period of the most frequent “outing” occurrence period, ”Is set as a period before the start of the start time zone of the occurrence time zone.

  Next, with reference to FIG. 33, processing for selecting information to be provided and output timing will be described. FIG. 33 is a flowchart illustrating an example of processing for selecting information to be provided and output timing.

  First, for a home 10 to be analyzed, a behavior tendency label labeled based on the estimation result of the living behavior of the home 10 is designated (step S401). This behavior trend label may be specified by an analyst. Further, the behavior tendency estimation unit 307 may specify the behavior tendency label labeled for the home 10 when the analyst selects the home 10.

  Next, in steps S402 to S405, the behavior tendency reflection recommendation information output unit 401 performs the processing of steps S403 to S404 for the number of defined home feature IDs. When the processing for the number of home feature IDs is completed, the behavior tendency reflection recommendation information output unit 401 proceeds to the process of step S406.

  First, in step S <b> 403, the behavior tendency reflection recommendation information output unit 401 sequentially determines the behavior trend labels of the home feature IDs to match the behavior trend labels of the designated home 10. If it is determined in step S403 that the action trend label of the designated home 10 is matched (YES), the action trend reflection recommendation information output unit 401 holds all the matched home feature IDs (step S404). . On the other hand, if it is determined in step S403 that the behavior trend label of the designated home 10 does not match (NO), the behavior trend reflection recommendation information output unit 401 returns the processing to step S403, and the next home feature ID. The behavior tendency label is determined.

  When the processing for the number of home feature IDs is completed, the behavior tendency reflection recommendation information output unit 401 proceeds to the process of step S406 (step S405).

  In step S406, the behavior tendency reflection recommendation information output unit 401 determines whether there is a matched home feature ID. If it is determined in step S406 that there is no matched home feature ID (NO), the behavior tendency reflection recommendation information output unit 401 ends the process without distributing the information.

  On the other hand, if it is determined in step S406 that there is a matching home feature ID (YES), the behavior tendency reflection recommendation information output unit 401 determines the recommended genre and the recommended advertising content according to the selection probability of the home feature ID. Is selected (see FIG. 31). At this time, the product of the selection probabilities may be further multiplied by a weighting factor (such as a random value of 0 to 1) and scored (step S407).

  Subsequently, the behavior tendency reflection recommendation information output unit 401 selects the recommended advertisement content and the transmission timing within a day so as to be equal to or less than a predetermined upper limit number (see step S408, FIG. 32). Here, the predetermined upper limit number is, for example, the upper limit number of contents that can be displayed.

  Then, the behavior tendency reflection recommendation information output unit 401 outputs the selected advertisement content at the selected transmission timing, and distributes it to the home 10 (step S409). The behavior tendency reflection recommendation information output unit 401 outputs not only the advertising content but also information based on a message or behavior tendency label (that is, information indicating a tendency of living behavior) as information to be provided to the user. May be.

  Thereby, the information terminal 15 of the home 10 acquires and displays information based on the advertising content, the message, or the behavior tendency label transmitted from the server 30 as information to be provided to the user.

  FIG. 34 is a diagram illustrating a display example of the display unit of the information terminal 15. In the display example shown in this figure, the total amount of power currently used in the home 10 (symbol D10), the amount of power used this month (symbol D20), the current amount of power used for each room or home appliance 13 (symbol D30). ) Etc. are displayed. In addition, information based on the analysis result analyzed by the behavior analysis device 32 is displayed (reference numeral D40). Here, information (symbol D41), message (symbol D42), and advertisement content (symbol D43) based on the behavior tendency label are displayed.

  As described above, the behavior analysis device 32 according to the first embodiment is a feature extraction unit 305 between occurrence times as a behavior feature extraction unit that extracts features of temporal changes in specific living behavior over a plurality of days. And an action sequence occurrence time zone feature extraction unit 306. In addition, the behavior analysis device 32 includes a behavior tendency estimation unit 307 that estimates a tendency of the lifestyle behavior based on the characteristic of temporal change of the particular lifestyle behavior. Thereby, the behavior analysis apparatus 32 can estimate a normal lifestyle appropriately. Moreover, the behavior analysis apparatus 32 can provide information suitable for a lifestyle to a user who is living the lifestyle.

(Second Embodiment)
Next, a behavior analysis system according to the second embodiment will be described.
The behavior analysis system according to the second embodiment differs from the behavior analysis system 1 shown in FIG. 2 in that a behavior analysis device 32A is provided instead of the behavior analysis device 32. The configuration other than the behavior analysis device 32A is the same as that of the behavior analysis system 1 shown in FIG.

  FIG. 35 is a block diagram illustrating an example of a schematic configuration of a behavior analysis apparatus 32A according to the second embodiment. In this figure, components corresponding to those in FIG. 2 are denoted by the same reference numerals, and description thereof is omitted. The behavior analysis device 32A includes an input unit 100A, a storage unit 200A, a calculation unit 300A, and an output unit 400.

  The input unit 100A includes an action label estimation condition input unit 102, an action sequence occurrence time zone extraction condition input unit 103, an occurrence time zone feature extraction condition input unit 104, and an action tendency estimation condition input unit 105. Yes. That is, the input unit 100A is different from the input unit 100 shown in FIG. 2 in that it does not include the electrical product operating state estimation condition input unit 101.

  The storage unit 200A includes an electrical product operation history data storage unit 202, a learning action record data storage unit 206, an action label sequence data storage unit 207, an action sequence occurrence time zone data storage unit 208, and action occurrence time zone feature data. A storage unit 209 and a recommended information storage DB 210 are provided. That is, the storage unit 200A has an electrical product power measurement data storage unit 201, an electrical product state history data storage unit 203, a human sensor measurement data storage unit 204, The difference is that the illuminance sensor measurement data storage unit 205 is not provided.

  The calculation unit 300A includes an action label estimation unit 303A, an action sequence occurrence time zone extraction unit 304, an occurrence time zone feature extraction unit 305, an action sequence occurrence time zone feature extraction unit 306, and an action tendency estimation unit 307. It is equipped with. That is, the calculation unit 300A is different from the calculation unit 300 illustrated in FIG. 2 in that the calculation unit 300A does not include the electric product operating state estimation unit 301 and the electric product state history collection unit 302. Further, the function of the behavior label estimation unit 303A is different from the behavior label estimation unit 303 shown in FIG.

  Based on the operation history data stored in the electrical product operation history data storage unit 202 and the learning action record data stored in the learning action record data storage unit 206, the action label estimation unit 303A Is estimated. Note that the action label used for estimation and the generation condition of the action label are input to the action label estimation condition input unit 102 and set.

  As described above, the behavior analysis apparatus 32A according to the second embodiment estimates the living behavior using the operation history data of each electrical product. Thereby, the behavior analysis apparatus 32A can appropriately estimate the lifestyle without using the measurement data of the electric energy of each electrical product in the home 10, the human sensor measurement data, and the temperature and humidity illuminance measurement data. . Therefore, the behavior analysis device 32A can estimate the lifestyle at a lower cost than the first embodiment without installing a device for measuring the amount of electric power in the home 10, various sensors 14, and the like.

  In the first embodiment, the behavior analysis device 32 estimates the living behavior using the power amount measurement data, operation history data, human sensor measurement data, and temperature / humidity illuminance measurement data of each electrical product. An example was explained. In the second embodiment, the example in which the behavior analysis device 32A estimates the living behavior using the operation history data of each electrical product has been described. However, the method of estimating the living behavior is not limited to these. For example, instead of using operation history data, human sensor measurement data, and temperature / humidity illuminance measurement data of each electrical product, it is possible to use a method for estimating living behavior using measurement data of electric energy of each electrical product. Moreover, it is good also as a method of estimating a living action using the measurement data of the electric energy of each electric product, the human sensor measurement data, and the temperature / humidity illumination measurement data without using the operation history data of each electric product.

  Moreover, in the said 1st, 2nd embodiment, although the home 10 was demonstrated as an analysis object of a living action, it is not restricted to this. Households, families, apartment houses, companies, organizations, etc. may be analyzed for living behavior.

  Further, the storage unit 200 illustrated in FIG. 2 and the storage unit 200A illustrated in FIG. 35 may not be included in the behavior analysis device 32 and the behavior analysis device 32A, respectively. For example, the storage unit 200 (200A) may be provided as a configuration other than the behavior analysis device 32 (32A) in the server 30. The storage unit 200 (200A) may be provided outside the server 30. In that case, the behavior analysis apparatus 32 (32A) is connected to the storage unit 200 (200A) provided outside via the network 20, and is stored in the storage unit 200 (200A) as in the case where it is provided inside. Can be used.

  In addition, although the behavior analysis device 32 illustrated in FIG. 2 and the behavior analysis device 32A illustrated in FIG. 35 have been described as including the output unit 400, the output unit 400 may be provided outside. That is, the behavior analysis device 32 and the behavior analysis device 32 </ b> A may be configured as devices until the tendency of the living behavior is estimated. Then, provision of information according to the tendency of the living behavior estimated by the behavior analysis device 32 and the behavior analysis device 32A may be performed by a device other than the behavior analysis device 32 and the behavior analysis device 32A.

  According to at least one embodiment described above, since the tendency of the living behavior is estimated based on the characteristic of the temporal change of the specific living behavior, the lifestyle can be appropriately estimated.

  You may make it implement | achieve the function of the action analysis apparatus 32 (32A) in embodiment mentioned above with a computer. In that case, a program for realizing this function may be recorded on a computer-readable recording medium, and the program recorded on this recording medium may be read into a computer system and executed. Here, the “computer system” includes an OS and hardware such as peripheral devices. The “computer-readable recording medium” refers to a storage device such as a flexible medium, a magneto-optical disk, a portable medium such as a ROM and a CD-ROM, and a hard disk incorporated in a computer system. Furthermore, the “computer-readable recording medium” dynamically holds a program for a short time like a communication line when transmitting a program via a network such as the Internet or a communication line such as a telephone line. In this case, a volatile memory inside a computer system serving as a server or a client in that case may be included and a program held for a certain period of time. The program may be a program for realizing a part of the functions described above, and may be a program capable of realizing the functions described above in combination with a program already recorded in a computer system.

  Although several embodiments of the present invention have been described, these embodiments are presented by way of example and are not intended to limit the scope of the invention. These embodiments can be implemented in various other forms, and various omissions, replacements, and changes can be made without departing from the spirit of the invention. These embodiments and their modifications are included in the scope and gist of the invention, and are also included in the invention described in the claims and the equivalents thereof.

DESCRIPTION OF SYMBOLS 1 ... Behavior analysis system, 11 ... Communication apparatus, 12 ... Distribution board, 13 ... Home appliance, 13a ... Electric kettle, 13b ... Microwave oven, 13c ... Rice cooker, 14 ... Sensor, 14a ... Human sensor, 14b ... Temperature Humidity illuminance sensor, 15 ... information terminal, 20 ... network, 30 ... server, 31 ... communication device, 32, 32A ... behavior analysis device, 100, 100A ... input unit, 101 ... electrical product operating state estimation condition input unit, 102 ... Action label estimation condition input unit, 103 ... Action sequence occurrence time zone extraction condition input unit, 104 ... Generation time zone feature extraction condition input unit, 105 ... Action trend estimation condition input unit, 200, 200A ... Storage unit, 201 ... Electricity Product power measurement data storage unit, 202 ... Electrical product operation history data storage unit, 203 ... Electrical product state history data storage unit, 204 ... Human sensor measurement data storage 205 ... Temperature / humidity illuminance sensor measurement data storage unit, 206 ... Learning action record data storage unit, 207 ... Action label sequence data storage unit, 208 ... Action sequence generation time zone data storage unit, 209 ... Action generation time zone feature data Storage unit 210 ... Recommended information storage DB 300, 300A ... Calculation unit 301 ... Electrical product operating state estimation unit 302 ... Electrical product state history collection unit 303, 303A ... Behavior label estimation unit 304 ... Behavior sequence occurrence time Band extraction unit, 305... Generation time zone feature extraction unit, 306... Action sequence generation time zone feature extraction unit, 307... Behavior trend estimation unit, 400.

Claims (13)

An action feature extraction unit that extracts characteristics of temporal changes of the specific living behavior over a plurality of days based on the presence or absence of specific living behavior by time period on multiple days;
A behavior tendency estimating unit for estimating a tendency of the specific living behavior based on the feature extracted by the behavior feature extracting unit;
Equipped with a,
A behavior estimation unit that estimates living behavior based on information related to electrical equipment;
Based on the specific living behavior estimated by the behavior estimating unit, an action occurrence time zone extracting unit that extracts an occurrence time zone of the specific living behavior among preset analysis target time zones;
With
The behavior feature extraction unit
Based on the occurrence time zone of the specific living behavior in a plurality of days extracted by the behavior occurrence time zone extraction unit, to extract the characteristics of the temporal change of the specific living behavior in the analysis target time zone,
The behavior estimation unit
Express life behavior to be estimated as action labels, and output action label series data in which one or more action labels are arranged in time series on a daily basis.
The action occurrence time zone extraction unit
With reference to the action label series data, an action label representing the specific living action is extracted from the action label series data including the action label, and the frequency of the action label extracted in the analysis target time zone and a predetermined number based of on the threshold, we extracted generation time period of the specific living activities in the analyzed time periods behavior analysis apparatus. An action feature extraction unit that extracts characteristics of temporal changes of the specific living behavior over a plurality of days based on the presence or absence of specific living behavior by time period on multiple days;
A behavior tendency estimating unit for estimating a tendency of the specific living behavior based on the feature extracted by the behavior feature extracting unit;
Equipped with a,
A behavior estimation unit that estimates living behavior based on information related to electrical equipment;
Based on the specific living behavior estimated by the behavior estimating unit, an action occurrence time zone extracting unit that extracts an occurrence time zone of the specific living behavior among preset analysis target time zones;
With
The behavior feature extraction unit
Based on the occurrence time zone of the specific living behavior in a plurality of days extracted by the behavior occurrence time zone extraction unit, to extract the characteristics of the temporal change of the specific living behavior in the analysis target time zone,
The behavior feature extraction unit
Based on the occurrence time zone of the specific living behavior in a plurality of days extracted by the behavior occurrence time zone extraction unit, classify each of the specific living behavior in a plurality of days for each combination that becomes the same occurrence time zone, the specific living behavior analyzer that to extract features of the temporal change in behavior in the analyzed time periods based on the classification result. An output unit that outputs information to be provided to a user related to the specific living behavior based on the tendency of the specific living behavior estimated by the behavior tendency estimating unit;
The behavior analysis device according to claim 1 or 2 . Information related to the electrical device includes information related to the operation history of the electrical device, or information related to power consumed by the electrical device.
The behavior analysis device according to any one of claims 1 to 3 . The behavior estimation unit
Furthermore, the living behavior is estimated based on information indicating the presence or absence of a person in the place where the living behavior is performed, or information indicating the temperature, humidity, or illuminance of the place,
The behavior analysis device according to any one of claims 1 to 4 . The behavior estimation unit
Information related to the electrical equipment, information indicating the presence or absence of a person at a place where the living action is performed, or information indicating the temperature or humidity of the place, and a living action record in which the actual living action for each time zone is recorded Based on the data, the teacher learns about the estimation condition of the living behavior across a plurality of time zones, and estimates the living behavior across the plurality of time zones based on the estimation condition.
The behavior analysis apparatus according to any one of claims 1 to 5 . The behavior estimation unit
The estimated daily behavior is expressed as an action label, and one or more action labels are arranged in time series in one unit from the beginning of the action label indicating the sleep action to immediately before the action label indicating the next sleep action. Output action label series data as an estimation result,
The behavioral analysis apparatus according to any one of claims 1 to 6, wherein The behavior feature extraction unit
Based on the occurrence time zone of the specific living behavior on a plurality of days extracted by the behavior occurrence time zone extraction unit, the start time, end time, and duration of each specific living behavior on a plurality of days Calculating one of average and variance, and frequency occupancy of a high frequency period as a characteristic of temporal change of the specific living behavior,
The behavior analysis device according to any one of claims 1 to 7 . The behavior tendency estimating unit
Priority of information provided to a user related to the specific living behavior among the plurality of candidates with reference to a storage unit storing a plurality of candidates of information provided to the user related to the specific living behavior Determining the degree based on the estimated tendency of the specific living behavior,
The behavior analysis device according to any one of claims 1 to 8 . The output unit is
Outputting information to be provided to the user related to the specific living behavior at a time specified based on the occurrence time zone of the specific living behavior;
The behavior analysis apparatus according to claim 3 . A behavior analysis method performed by a computer,
Extracting characteristics of temporal changes of the specific living behavior over a plurality of days based on the presence or absence of the specific living behavior for each time zone in a plurality of days; and
Estimating a tendency of the specific living behavior based on the extracted feature;
It includes,
Estimating life behavior based on information related to electrical equipment;
Based on the specific living behavior estimated by the step of estimating the living behavior, extracting an occurrence time zone of the specific living behavior from a preset analysis target time zone;
Further including
In the step of extracting the characteristic of temporal change of the specific living behavior, the time of the specific living behavior in the analysis target time zone based on the occurrence time zone of the specific living behavior in a plurality of extracted days The characteristics of typical changes,
In the step of estimating the living behavior, the estimated living behavior is represented as an action label, and action label series data in which one or a plurality of action labels are arranged in a time series on a daily basis is output as an estimation result,
In the step of extracting the occurrence time zone of the specific living action, referring to the action label series data, extracting an action label representing the specific living action from the action label series data including the action label The behavior analysis method for extracting the occurrence time zone of the specific living behavior in the analysis target time zone based on the frequency of the action label extracted in the analysis target time zone and a predetermined threshold . A behavior analysis method performed by a computer,
Extracting characteristics of temporal changes of the specific living behavior over a plurality of days based on the presence or absence of the specific living behavior for each time zone in a plurality of days; and
Estimating a tendency of the specific living behavior based on the extracted feature;
It includes,
Estimating life behavior based on information related to electrical equipment;
Based on the specific living behavior estimated by the step of estimating the living behavior, extracting an occurrence time zone of the specific living behavior from a preset analysis target time zone;
Further including
In the step of extracting the characteristic of temporal change of the specific living behavior, the time of the specific living behavior in the analysis target time zone based on the occurrence time zone of the specific living behavior in a plurality of extracted days The characteristics of typical changes,
In the step of extracting the characteristic of temporal change of the specific living behavior, each of the specific living behaviors on a plurality of days is the same based on the time period of occurrence of the specific living behavior on the extracted plurality of days. classified by combination of a generation time period, behavior analysis how to extract a characteristic of the temporal change of the specific living activities in the analyzed time periods based on the classification result. A program for causing a computer to execute the behavior analysis method according to claim 11 or 12 .

Images