JP7797261B2 - Learning system and learning method - Google Patents

Description

The present invention relates to a learning system and learning method technology that learns the results of detecting a subject's behavior.

Traditionally, learning systems and methods for learning the results of detecting the behavior of people residing in a building have been publicly known. For example, see Patent Document 1.

Patent Document 1 describes a lifestyle pattern management system that uses the amount of energy used in a building to detect changes in the lifestyle patterns (behavior) of residents.

When detecting a resident's lifestyle patterns using energy usage, as in the invention described in Patent Document 1, even if energy usage data is acquired, it is not possible to effectively detect the resident's lifestyle patterns when the resident's lifestyle is not stable, such as when the resident first moves in. As a result, unnecessary data may be acquired until the resident's lifestyle stabilizes, and there is a risk that a lifestyle pattern that reflects the resident's actual situation will not be detected.

JP 2016-48503 A

The present invention was made in light of the above situation, and the problem it aims to solve is to provide a learning system and learning method that can create learning behavior data that is in line with the actual behavioral patterns of a subject.

The above is the problem that this invention aims to solve, and next we will explain the means for solving this problem.

That is, in claim 1, the system comprises a behavior detection unit capable of detecting the behavior of the subject by continuously detecting the usage status of equipment installed in a building used by the subject, a data creation unit that creates learned behavior data regarding the behavior of the subject by learning the detection results by the behavior detection unit, and a condition setting unit that sets conditions for the creation of the learned behavior data by the data creation unit so that the learned behavior data does not deviate from the actual behavior pattern of the subject, wherein the data creation unit is capable of creating actual behavior data regarding the behavior of the subject based on the detection results by the behavior detection unit, is capable of evaluating a comparison result of the actual behavior data with the learned behavior data, and is capable of evaluating the comparison result of the actual behavior data with the learned behavior data for at least two or more multiple cycles having different periods, calculates an average value regarding the use of the equipment for each cycle, and performs evaluation for each of the multiple cycles in which the calculated average value satisfies a predetermined threshold .

In claim 2, the condition setting unit is configured to cause the data creation unit to start learning when the behavior detection unit detects the use of the equipment for the first time, as one of the conditions.

In claim 3, when the comparison result is evaluated as being outside a predetermined tolerance range as one of the conditions, the condition setting unit executes a learning necessity determination process to determine whether or not learning is necessary for the corresponding detection result, which is the detection result corresponding to the executive movement data.

In claim 4, the condition setting unit, in the learning necessity determination process, notifies a predetermined notification unit of a request for feedback regarding the need for learning of the correspondence detection result, and determines whether the correspondence detection result needs to be learned based on the acquired feedback.

In claim 5, when the correspondence detection result indicates that learning is not necessary during the learning necessity determination process, the condition setting unit determines whether to execute the learning necessity determination process the next time a detection result with the same content as the correspondence detection result is detected.

In a sixth aspect of the present invention, the equipment includes a plurality of home appliances, and the condition setting unit sets the condition for each of the plurality of home appliances .

The effects of the present invention are as follows:

In this invention, it is possible to create learning behavior data that is in line with the actual behavioral patterns of the subject.

1 is a schematic diagram showing the configuration of a lifestyle pattern determination system according to an embodiment of the present invention; FIG. 1 is a diagram showing an overview of a series of processes performed by a lifestyle pattern determination system. 10A to 10C are diagrams showing examples of learning results, detection results in daily processing, and determination results. 10 is a flowchart showing the contents of a learning start determination process. 10 is a flowchart showing the contents of a home appliance use count evaluation process. 10 is a flowchart showing the contents of a learning data update necessity determination process. 10 is a flowchart showing the contents of an evaluation period determination process.

The following describes the configuration of a lifestyle pattern determination system 100 according to one embodiment of the present invention, using Figure 1.

The lifestyle pattern determination system 100 learns the lifestyle (behavioral) patterns of subject P1 and determines whether or not there are any abnormalities in the lifestyle patterns based on the learning results. In this embodiment, as an example, subject P1 is an elderly person residing in residence 1, and the presence or absence of abnormalities in subject P1's lifestyle patterns (and therefore the presence or absence of abnormalities in subject P1's cognitive function) is assumed. The lifestyle pattern determination system 100 mainly comprises a power sensor 110, a server 120, and a terminal 130.

The power sensor 110 detects the power consumption of various pieces of equipment used by the subject P1 (particularly equipment used in daily life). The power sensor 110 is installed in the distribution board 2 of the house 1 in which the subject P1 resides. The power sensor 110 can detect the power consumption of each branch circuit of the distribution board 2. This allows the power sensor 110 to detect the usage status (whether or not the equipment is being used) of the equipment connected to each branch circuit. By detecting the usage status of the equipment, it is possible to indirectly detect the behavior of the subject P1 (which equipment is being used). The power sensor 110 is installed in the house 1, for example, when the lifestyle pattern determination system 100 is introduced, and performs detection at all times (continuously) thereafter.

In this embodiment, examples of equipment for which power consumption is detected by the power sensor 110 include home appliances such as cooking appliances (induction cookers, microwave ovens, etc.), refrigerators, televisions, hair dryers, vacuum cleaners, washing machines, air conditioners, heaters, and lighting. For example, if use of a cooking appliance is detected, it is indirectly detected that subject P1 has used the cooking appliance (and therefore eaten). Furthermore, if use of a television is detected, it is indirectly detected that subject P1 has watched television.

Note that the method of detecting the usage status of various pieces of equipment is not limited to detecting the power of each branch circuit using the power sensor 110, and various other methods can be used. For example, it is possible to detect the power (main power) of the main circuit of the distribution board 2 and analyze the waveform of that power to identify and understand the equipment being used. Also, rather than detecting the power of the distribution board 2, it is possible to directly detect the operating status of the various pieces of equipment themselves (such as whether the power of the various pieces of equipment is on or off, the power of the outlets to which the various pieces of equipment are connected, etc.). Furthermore, if the home 1 is equipped with a system that monitors (manages) the usage status of various pieces of equipment (such as a HEMS: Home Energy Management System), it is also possible to use the information monitored by that system.

The server 120 performs various processes based on the detection results of the power sensor 110. The server 120 is configured, for example, as a virtual server (cloud server) installed on the cloud. By acquiring information from the power sensor 110, the server 120 can grasp the usage status of various pieces of equipment. The server 120 learns the usage status (measurement data) of various pieces of equipment and can generate a lifestyle pattern (learned behavior pattern) of the subject P1 since the introduction of the system. The server 120 can also generate a lifestyle pattern (executive behavior data) of the subject P1 on the evaluation date based on newly acquired measurement data. In this way, the server 120 can determine whether or not there is an abnormality in the lifestyle pattern of the subject P1 (and therefore an abnormality in cognitive function) by comparing the generated learned behavior pattern with the executive behavior data. The server 120 can also send and receive various information to and from the terminal 130, which will be described later.

The terminal 130 is capable of displaying various types of information. The terminal 130 is carried by a person (e.g., a family member or relative of the subject P1) who should be aware of whether or not there are any abnormalities in the lifestyle patterns of the subject P1. In this embodiment, the terminal 130 is carried by a family member P2 of the subject P1. The terminal 130 is configured, for example, as a device (e.g., a smartphone or tablet device) that can be carried by the family member P2 of the subject P1. The terminal 130 can notify the family member P2 of the subject P1 of information from the server 120 in an appropriate manner (e.g., display on an LCD screen, audio, etc.).

By using the lifestyle pattern determination system 100 configured as described above, it is possible to detect signs of or the onset of dementia in subject P1 based on changes in the subject P1's lifestyle pattern.

For example, symptoms of dementia may include memory impairment, disorientation, executive dysfunction, and a reversal of day and night cycles.

"Memory impairment" is a disorder that causes symptoms such as an inability to remember new things and a loss of previously remembered memories. When subject P1 develops "memory impairment," it is expected that behavioral changes (abnormalities) will occur, such as forgetting that they have eaten and eating again, or forgetting that they have cleaned and cleaning again.

"Disorientation" is a disorder in which a person is unable to grasp the situation they are in, such as "when, where, who," etc. If subject P1 develops "disorientation," it is expected that they will experience behavioral changes (abnormalities) such as turning on the heater even though it is summer, becoming unable to find their way around the house and being unable to go to the toilet or bathroom, or being unable to return home once they have gone out.

"Executive dysfunction" is a disorder that makes it difficult to plan and carry out things in an orderly manner. If subject P1 were to develop "executive dysfunction," it is expected that behavioral changes (abnormalities) would occur, such as being unable to prepare meals or not knowing how to use electrical appliances.

"Day-night reversal" is a disorder in which the sleep-wake rhythm is disrupted and day and night are reversed. When subject P1 develops "day-night reversal," it is expected that behavioral changes (abnormalities) will occur, such as being active during the time when they should be sleeping (in the middle of the night).

The lifestyle pattern determination system 100 of this embodiment therefore detects changes in the behavior (lifestyle pattern) of the subject P1 as described above, and determines whether or not the subject P1 has an abnormality in cognitive function based on this change in behavior. If necessary, the system then notifies the subject P1's family P2 of the abnormality in the subject P1's cognitive function. This allows the subject P1's family P2 to recognize the signs and onset of dementia in the subject P1, and to take appropriate measures (treatment, etc.) early on.

The following provides an overview of the series of processes performed by this lifestyle pattern determination system 100.

As shown in FIG. 2, the server 120 stores various pieces of equipment installed in the residence 1. In this embodiment, the server 120 stores a plurality of home appliances as various pieces of equipment, such as cooking appliances (induction cooker, microwave), refrigerator, television, hair dryer, vacuum cleaner, washing machine, and seasonal appliances (air conditioner, heating appliance). The server 120 performs appropriate processing for each of these home appliances to determine whether or not the subject P1 has an abnormality in cognitive function, and notifies the subject P1's family member P2 of the abnormality in cognitive function as necessary.

Specifically, the server 120 detects daily usage (and thus the behavior of the subject P1) and performs processing to determine whether or not there is an abnormality in the cognitive function of the subject P1 based on the detection results. This processing is hereinafter referred to as "daily processing." In daily processing, the server 120 stores the behavior that was the subject of the judgment, whether or not there is an abnormality, and the reason for the judgment, linking them together.

The server 120 also detects the usage status (and thus the behavior of the subject P1) on a weekly (weekly) basis, and performs a process to determine whether or not the subject P1 has an abnormality in their cognitive function based on the detection results. Hereinafter, this process will be referred to as "weekly processing." In the weekly processing, the server 120 stores the behavior that was the subject of the assessment, whether or not there is an abnormality, and the reason for the assessment, linking them together.

The server 120 also detects monthly (weekly) usage (and thus the behavior of the subject P1) and performs processing to determine whether or not there is an abnormality in the cognitive function of the subject P1 based on the detection results. Hereinafter, this processing will be referred to as "monthly processing." In the monthly processing, the server 120 stores the behavior that was the subject of the judgment, whether or not there is an abnormality, and the reason for the judgment, linking them together.

The server 120 also calculates the number of times an abnormality is detected for each behavior during the week through the daily processing described above. Then, for behaviors for which an abnormality is detected a predetermined number of times or more, an alert (warning) is issued using the terminal 130. Hereinafter, this processing will be referred to as "weekly result notification." The weekly result notification enables the subject P1's family member P2, who is carrying the terminal 130, to understand that an abnormality has occurred in the subject P1's specified behavior, and ultimately that there is a risk of cognitive function decline.

The server 120 also calculates the number of times an abnormality is detected for each behavior during a month through the weekly and monthly processing. Then, for behaviors for which an abnormality is detected a predetermined number of times or more, an alert is issued using the terminal 130. Hereinafter, this process will be referred to as "monthly result notification." This allows the subject P1's family member P2, who owns the terminal 130, to understand that an abnormality has occurred in the subject P1's behavior, and ultimately that there is a risk of cognitive function decline.

In this way, the lifestyle pattern determination system 100 performs each process for each target home appliance (target home appliance) to determine whether or not there is an abnormality in the cognitive function of the target person P1 and can issue an alert to the target person P1's family member P2. Whether daily, weekly, or monthly processing is performed for each target home appliance is determined by executing the evaluation period determination process described below.

As a prerequisite for the process of determining whether or not subject P1 has an abnormality in cognitive function, server 120 detects the usage status of various facilities in home 1 over a predetermined period of time and stores the information in a predetermined storage area. In this way, server 120 acquires the detection results (measurement data) of the power sensor 110 and, by executing the data management process described below, appropriately learns the acquired measurement data (i.e., subject P1's behavioral tendencies) and generates the learning results (learned behavior data). As a result, server 120 detects the actions and time periods taken by subject P1 throughout their daily lives, and understands the actions taken by subject P1 for each time period. In other words, the learned behavior data includes the time periods and number of times per day that subject P1 has performed the actions since the system was introduced. The "learning results" in the table in Figure 3 shows an example of learned behavior data.

For example, when the server 120 detects that a cooking appliance (induction cooker, microwave) has been used, it learns the time of day. This is repeated for a predetermined period of time to understand the behavior of the subject P1, including the time of day the subject P1 usually uses the cooking appliance. The server 120 may also indirectly detect that the subject P1 has gone to bed by, for example, turning off the lights in the house 1, and learn the sleeping hours of the subject P1.

Note that the behavior of subject P1 changes depending on the season, such as air conditioners and heating appliances being used only in seasons when they are needed (summer and winter). Therefore, server 120 learns the behavioral trends of subject P1 for each season, and when performing daily processing, etc., as described below, it can also use the learning results according to the season at that time.

Furthermore, during the period that is the subject (target) of processing (e.g., daily processing) such as the above, the server 120 constantly detects the usage status of the various facilities in the house 1 and stores the information in a specified memory area. In this way, the server 120 acquires the detection results (measurement data) of the power sensor 110 and generates the usage status of the various facilities in the house 1 (i.e., the actual behavior data of the subject P1) based on the acquired measurement data. In other words, the actual behavior data includes the time periods of the activities performed by the subject P1 during the period that is the target of evaluation, the number of times per specified period, etc. The server 120 can also update the learned behavior data as appropriate based on the actual behavior data (or newly acquired measurement data). The "Detection Results" in the table in Figure 3 shows an example of actual behavior data.

In this way, the server 120 can use the learned behavior data to ascertain the criteria (time period, number of occurrences, etc.) for the behaviors that are the subject of judgment (target of judgment) for the presence or absence of abnormalities. The server 120 then compares the reference learned behavior data with the actual movement data for the period to be judged, thereby determining whether or not there is an abnormality in the lifestyle pattern of subject P1 (and therefore whether or not there is an abnormality in cognitive function).

For example, when server 120 determines whether or not there is an abnormality based on the number of times a target appliance is used each day, it determines (sets) an acceptable range for the number of times the target appliance is used using a predetermined probability distribution from the learned behavior data. Next, server 120 determines whether the number of times the target appliance is used, included in the execution behavior data, is within an acceptable range (predetermined threshold), and if the number of times it is used is within the acceptable range, it determines that it is normal (performs a normal determination), and if it is outside the acceptable range, it determines that it is abnormal (performs an abnormality determination).

Furthermore, when determining whether or not there is an abnormality based on the time period of the target behavior, the server 120 determines whether or not the time period of use of the target appliance in the detection results (actual behavior data) matches the time period of use of the target appliance in the learning results (learned behavior pattern), as shown in Figure 3. In this way, the server 120 determines that there is an abnormality if the time period of use of the target appliance in the detection results (actual behavior data) matches the time period of use of the target appliance in the learning results (learned behavior pattern), and determines that there is an abnormality if they do not match.

The following describes the data management process performed by server 120.

The data management process appropriately manages the acquired measurement data so that the created learning behavior data does not deviate from the actual lifestyle patterns of subject P1. The data management process includes a learning start determination process, a frequency evaluation/update necessity determination process, and an evaluation period determination process.

First, the learning start determination process performed by server 120 will be explained below using Figure 4.

For example, when subject P1 moves into house 1, all of the target equipment (various types of equipment) may not be installed. In such a case, if learning is started using measurement data acquired immediately after subject P1 moves in, learning behavior data will be generated based on measurement data obtained when the target appliance (target appliance) is not present. In this way, target appliances installed some time after subject P1 moves in will be determined to be target appliances with low usage frequency at the time of installation, regardless of their actual usage frequency, which creates a problem in that the learning behavior data may not match the actual situation.

In response to this, the lifestyle pattern determination system 100 executes a learning start determination process, and the timing for starting learning is determined by the server 120 rather than at the time of moving in, thereby preventing the occurrence of the above-mentioned problems. The learning start determination process is executed, for example, once a day (e.g., at midnight) based on measurement data from the previous 24 hours.

As shown in FIG. 4, in the learning start determination process, the server 120 repeats steps S101 to S106 for each target appliance of each type of equipment. The following describes each step from S101 to S106.

First, in step S101, the server 120 reads 24 hours' worth of measurement data for the target appliance stored in a specified storage area. Next, in step S102, the server 120 determines whether the detection flag for the target appliance is OFF.

Here, the detection flag indicates information on whether the power consumption of the target appliance has been detected in the past (i.e., after the lifestyle pattern determination system 100 was installed in the home 1). In other words, if the detection flag is ON, it indicates that the power consumption of the target appliance has been detected in the past. On the other hand, if the detection flag is OFF, it indicates that the power consumption of the target appliance has not been detected in the past. The detection flag is set for each target appliance by the server 120.

If the detection flag is OFF (YES in step S102), the server 120 proceeds to step S103. If the detection flag is OFF, the power consumption of the target appliance has never been detected in the past, and therefore the target appliance may not yet be installed in the home 1. On the other hand, if the detection flag is ON (NO in step S102), the server 120 ends the learning start determination process for the target appliance because the power consumption of the target appliance has been detected in the past (i.e., learning has already started for the target appliance).

In step S103, the server 120 determines whether the power consumption of the target appliance is detected in the measurement data (the measurement data read in step S101). If the server 120 determines that the power consumption of the target appliance is detected in the measurement data (YES in step S103), it proceeds to step S104. On the other hand, if the server 120 determines that the power consumption of the target appliance is not detected in the measurement data (NO in step S103), it does not start learning for the target appliance and therefore ends the learning start determination process for the target appliance.

In step S104, the server 120 sets the detection flag for the target appliance from OFF to ON. Next, in step S105, the server 120 sets the measurement data acquired from now on, including the measurement data, for the target appliance as the learning target. In this way, the server 120 uses the measurement data acquired from now on for the target appliance to generate learning behavior data from the timing of the next learning (for example, the following Sunday).

Next, in step S106, the server 120 notifies the terminal 130 that it has detected the power consumption of the target appliance (i.e., that the target appliance has been used). Specifically, the server 120 adds a text indication such as "Induction cooking appliance use has begun" to the content of the weekly result notification sent via the terminal 130.

In this way, by executing the learning start determination process, learning begins when use of a target appliance is detected for the first time, so learning behavior data will be generated for appliances that have been used at least once in the past. In this way, it is possible to prevent learning behavior data from being generated for each target appliance based on measurement data acquired when, for example, a target appliance is not installed (when the target appliance is not present). In other words, learning can be started at the appropriate time for each target appliance, preventing the learning behavior data from mismatching (diverging) with reality.

Next, using Figures 5 and 6, we will explain the process performed by server 120 to evaluate the number of times and determine whether or not an update is necessary.

For example, seasonal appliances such as air conditioners and heaters are not used during the intermediate periods, resulting in variations in usage frequency throughout the year. Furthermore, for example, broken appliances (broken appliances) cannot be used even if one wishes to. Thus, if measurement data from periods when seasonal appliances or broken appliances are not used is learned, the learned behavioral data will not match the actual situation, such as an abnormality being detected despite the subject P1's normal (normal) lifestyle pattern, and so this should not be learned.

On the other hand, if an abnormality is determined to be due to a change in the lifestyle pattern of subject P1, for example, the measurement data for which this abnormality was determined should be learned (reflected in the learned behavior data). As such, even for measurement data for which the same abnormality was determined, there is a mixture of data that should be learned and data that should not be learned (measurement data from periods when seasonal appliances or faulty appliances are not in use), which poses a problem in that it is difficult to determine whether the measurement data needs to be learned.

The lifestyle pattern determination system 100 therefore performs a frequency evaluation and update necessity determination process, and while providing feedback from the subject P1, family P2, etc., determines whether the measurement data used to determine an abnormality should be learned, thereby preventing the occurrence of problems such as those described above. The frequency evaluation and update necessity determination process targets target appliances whose detection flags are determined to be ON in the learning start determination process.

The frequency evaluation and update necessity determination process includes an appliance usage frequency evaluation process and a learning data update necessity determination process. Specifically, for target appliances whose detection flag is determined to be ON in the learning start determination process, the appliance usage frequency evaluation process is first executed, and then, as necessary, the learning data update necessity determination process is executed.

First, we will explain the appliance usage count evaluation process, which is part of the count evaluation and update necessity determination process, using Figure 5.

As shown in FIG. 5, in the home appliance usage count evaluation process, the server 120 repeats the processes from step S201 to step S210 for each target home appliance of each type of equipment. The following describes each process from step S201 to step S210.

First, in step S201, the server 120 reads 24 hours' worth of measurement data for the target appliance stored in a specified storage area. Next, in step S202, the server 120 determines whether the interruption flag for the target appliance is OFF.

Here, the suspend flag indicates whether the measurement data is to be learned (reflected in the learned behavior data). In other words, if the suspend flag is ON, it indicates that the measurement data is not to be learned (reflected in the learned behavior data). If the suspend flag is OFF, it indicates that the measurement data is to be learned (reflected in the learned behavior data). The suspend flag is set for each target home appliance by the server 120.

If the interruption flag is OFF (YES in step S202), the server 120 proceeds to step S206. On the other hand, if the interruption flag is ON (NO in step S202), the server 120 proceeds to step S203.

In step S203, the server 120 determines whether the target appliance has been used. That is, the server 120 determines whether the power consumption of the target appliance has been detected in the measurement data. If the power consumption of the target appliance has been detected, the server 120 determines that the target appliance has been used and proceeds to step S204.

On the other hand, if the server 120 has not detected the power consumption of the target appliance, it determines that the target appliance has not been used and terminates the appliance use count evaluation process for that target appliance. In other words, if the interruption flag is ON (NO in step S202) (i.e., a period of non-use, such as a seasonal appliance or a malfunctioning appliance), and the target appliance has not been used (NO in step S203), it is determined that the non-use period continues, and the appliance use count evaluation process is terminated.

In step S204, the server 120 sets the interruption flag for the target appliance from ON to OFF. Next, in step S205, the server 120 notifies the terminal 130 that it has redetected the power consumption of the target appliance (i.e., that use of the target appliance has resumed). Specifically, the server 120 adds a text indication such as "Use of induction heater has resumed" to the content of the weekly result notification sent via the terminal 130.

In step S206, the server 120 evaluates the number of times the target appliance is used. Specifically, as described above, the server 120 sets an acceptable range for the number of times the target appliance is used from the learned behavior data, and then determines whether the number of times the target appliance is used, included in the performance behavior data, is within the acceptable range. In this way, the server 120 evaluates the results of comparing the performance behavior data with the learned behavior data.

Specifically, if the number of uses of the target appliance included in the execution behavior data is within the allowable range (if the comparison result is within the allowable range), the server 120 makes a normal judgment. On the other hand, if the number of uses of the target appliance included in the execution behavior data is outside the allowable range (if the comparison result is outside the allowable range), the server 120 makes an abnormal judgment. Note that when step S206 is transitioned from step S205, the learning behavior data of the target appliance used for the judgment is the data before the interruption.

Note that the method for evaluating the results of comparing the active behavior data is not limited to the above. Various methods can be used as long as they can evaluate the degree of difference between the active behavior data and the learned behavior data and determine whether the data is abnormal or normal.

Next, in step S207, the server 120 notifies the terminal 130 of the evaluation result (abnormal or normal) from step S206.

Next, in step S208, the server 120 determines whether the evaluation result in step S206 is an abnormality judgment. If the server 120 determines that the evaluation result in step S206 is an abnormality judgment (YES in step S208), it proceeds to step S209. On the other hand, if the server 120 determines that the evaluation result in step S206 is a normal judgment (NO in step S208), it proceeds to step S210.

In step S209, the server 120 sets the measurement data for which an abnormality has been determined as unlearned data. Here, unlearned data is data for which it has not yet been determined whether it should be learned or not. In other words, the server 120 temporarily suspends the decision on whether to learn the measurement data for which an abnormality has been determined, and terminates the home appliance usage frequency evaluation process. At this time, the server 120 temporarily stores the unlearned data in a specified storage area. Meanwhile, in step S210, the server 120 reflects the measurement data for which a normal determination has been made as being something that should be learned in the learned behavior data. In other words, the server 120 updates the learned behavior data.

Next, we will explain the process for determining whether learning data needs to be updated, which is part of the frequency evaluation and update necessity determination process, using Figure 6.

The process for determining whether or not to update learned data is executed for the target appliance from which unlearned data was acquired when unlearned data is temporarily stored in a specified storage area during the appliance usage frequency evaluation process.

In step S301, the server 120 uses the terminal 130 to notify, for example, a device portable by family member P2, that an abnormality has been determined, and requests feedback regarding whether or not the measurement data for which an abnormality has been determined is problematic (and therefore whether or not the measurement data needs to be learned).

Specifically, the server 120 displays on the notification app embedded in the device one of the following four options, prompting the user to select and respond: (a) "There is a problem (notify again if a similar abnormality occurs); (b) "No problem (it just so happened to be different from usual. Notify again if a similar abnormality occurs)); (c) "No problem (no problem as lifestyle patterns have changed)"; (d) "No problem (temporarily not in use due to seasonal appliance or faulty appliance)." In this way, the server 120 can receive feedback from family member P2's operation of the notification app.

The feedback period (period for accepting responses) is set, for example, until the next update date of the learned behavior pattern. In other words, if the learned behavior pattern is updated once a week (for example, on Sunday), the server 120 will accept feedback until the next update date (Sunday). If the server 120 does not receive feedback by the set date, it can discard the unlearned data stored in a specified area.

In step S302, the server 120 determines, based on the received feedback response, whether the abnormality determination is due to the fact that the target appliance is currently not in use. Specifically, the server 120 determines whether the feedback response is (d) or any one of (a), (b), and (c).

Thus, if the feedback response is (d), the server 120 determines that the abnormality determination was caused by the period during which the target appliance was not in use (YES in step S302) and proceeds to step S303. On the other hand, if the feedback response is any one of (a), (b), or (c), the server 120 determines that the abnormality determination was not caused by the period during which the target appliance was not in use (it was not a period during which the target appliance was not in use) (NO in step S302) and proceeds to step S305.

In step S303, server 120 discards the unlearned data stored in a specified area. Note that server 120 continues to retain the current learned behavior data (learning results up to now) for the target appliance in a specified storage area, in order to use it when the interruption ends and judgment resumes (after the end of the non-use period).

In step S304, the server 120 sets the interruption flag for the target appliance from OFF to ON. After step S304, the server 120 then terminates the process for determining whether or not the learning data needs to be updated for the target appliance.

In step S305, to which the server 120 proceeds after determining that it is not currently in a non-use period (NO in step S302), the server 120 determines whether to perform an abnormality determination again if data with the same content as the measurement data used for the abnormality determination is acquired next time or later. Specifically, if the feedback response is either (a) or (b), the server 120 determines that an abnormality determination will be performed again (YES in step S305) and proceeds to step S306. On the other hand, if the feedback response is (c), the server 120 determines that an abnormality determination will not be performed again (NO in step S305) and proceeds to step S307.

In step S306, the server 120 discards the unlearned data stored in the specified area. The server 120 then terminates the process for determining whether or not the learned data needs to be updated for the target appliance.

In step S307, the server 120 reflects the measurement data for which an abnormality has been determined to be present in the learned behavior data, since the measurement data should be learned. In other words, the server 120 updates the learned behavior data. The server 120 then terminates the process for determining whether or not the learned data needs to be updated for the target appliance.

In this way, by executing the frequency evaluation and update necessity determination process, even if the measurement data has been determined to be the same abnormality, it is possible to distinguish between what should be learned (for example, the measurement data for option (c) above) and what should not be learned (for example, the measurement data for options (a), (b), and (d) above), and the measurement data can be learned based on this distinction. In this way, the accuracy of the learning behavior data can be improved.

Similarly, even if measurement data is determined not to need to be learned, the next time the same measurement data is acquired, it is possible to distinguish between data that will be reassessed as abnormal and data that will not be reassessed as abnormal. This prevents the need to repeatedly request feedback from family member P2 of subject P1, thereby reducing the burden on family member P2.

For example, if subject P1 has a visitor, an abnormality may be detected by using an induction cooktop or other home appliance at an unusual time. Because such abnormality detections are not continuous, the family member providing the feedback, for example, P2, can choose to perform an abnormality detection (give feedback) the next time the same measurement data is acquired.

Next, we will explain the evaluation period determination process performed by server 120 using Figure 7.

Here, the frequency of use of the target appliances varies depending on the target person and the type of target appliance. For example, as shown in Figure 3, among the target appliances, a microwave oven is used three times a day, while a vacuum cleaner (not shown) is often used two or three times a week. Furthermore, even for the same target appliance (e.g., an air conditioner), the frequency of use will vary depending on the target person. Therefore, if the processing (evaluation) cycle for determining whether or not there is an abnormality in the lifestyle pattern is set uniformly, there is a problem in that the accuracy of the determination results may be low.

The lifestyle pattern determination system 100 therefore executes an evaluation cycle determination process to prevent the occurrence of problems such as those described above. Specifically, by executing the evaluation cycle determination process, it is possible to determine the evaluation cycle (the timing of evaluation) for each subject and each subject appliance. In this embodiment, as described above, three types of cycles can be used for evaluation: daily, weekly, and monthly. In other words, in this embodiment, the three types of evaluation cycles that can be determined are daily (day), weekly (week), and monthly (month).

Note that the server 120 is unable to determine the evaluation period in the initial stage (when measurement data has not yet been accumulated to the extent that it can be learned, such as immediately after the system is introduced), and is therefore pre-configured to perform daily processing (to perform evaluations on a daily basis). The evaluation period determination process is executed for each subject (subject P1 in this embodiment), each subject appliance, and each set evaluation period after measurement data has been accumulated to the extent that it can be learned.

In step S401, the server 120 calculates (updates) the average value for the usage of the target appliance for each evaluation period. Specifically, the server 120 calculates the average values (daily average value and weekly average value) of the number of times the target appliance is used for each day and week at the current time, excluding the longest evaluation period (every month). In this way, the server 120 calculates the average daily average value and weekly average value of the number of times the target appliance is used. Note that the average value for usage may not be the average value of the number of times of use as described above, but may instead be the average value of other criteria related to usage, such as usage time.

Next, in step S402, the server 120 determines whether the updated daily average value is below the minimum standard. Here, the minimum standard is a threshold for determining whether to perform daily processing. In this embodiment, the minimum standard is set to 1. If the updated daily average value is below the minimum standard (YES in step S402), the server 120 proceeds to step S403. On the other hand, if the updated daily average value is equal to or greater than the minimum standard (NO in step S402), the server 120 proceeds to step S406.

In step S403, the server 120 determines whether the updated weekly average value is below the minimum standard. Here, the minimum standard is a threshold for determining whether to perform weekly processing. In this embodiment, as in step S402, the minimum standard is set to 1. If the updated weekly average value is below the minimum standard (YES in step S403), the server 120 proceeds to step S404. On the other hand, if the updated weekly average value is equal to or greater than the minimum standard (YES in step S403), the server 120 proceeds to step S405.

In step S404, the server 120 determines to perform only monthly processing of the target appliances among daily, weekly, and monthly processing. That is, the evaluation period is set to once a month (month). Thus, if the updated daily average value is below the minimum standard and the updated weekly average value is below the minimum standard (YES in step S402, YES in step S403), that is, if the average number of uses per day and per week is less than 1, it is determined that it is difficult to perform daily and weekly evaluations, and therefore only monthly processing is performed. After executing step S404, the server 120 terminates the evaluation period determination process.

In step S405, the server 120 determines whether to perform weekly processing and monthly processing for the target appliance out of daily processing, weekly processing, and monthly processing. That is, weekly (week) and monthly (month) are set as the evaluation cycles. Thus, if the updated daily average value is below the minimum standard and the updated weekly average value is equal to or greater than the minimum standard (YES in step S402, NO in step S403), that is, if the average number of daily uses is less than 1 but the average number of weekly uses is 1 or greater, it is determined that daily evaluation is difficult, but weekly and monthly evaluation is possible, and therefore weekly and monthly processing is performed. After executing step S405, the server 120 terminates the evaluation cycle determination process.

In step S406, the server 120 determines to perform all of the following processes for the target appliance: daily, weekly, and monthly. That is, the evaluation periods are set to daily (day), weekly (week), and monthly (month). If the updated daily average value is equal to or greater than the minimum standard (NO in step S402), that is, if the average number of daily uses is 1 or greater, it is determined that daily evaluation, which is the shortest evaluation period, is possible, and all of the processes (daily, weekly, and monthly) are performed. After executing step S406, the server 120 terminates the evaluation period determination process.

In this way, by executing the evaluation period determination process, it is possible to determine the evaluation period (when to perform evaluation) for each target user and target appliance, thereby improving the accuracy of the determination results. Furthermore, for example, as shown in steps S404 and S405, if it is decided to perform evaluation at some of the multiple evaluation periods, there is no need to perform processing at periods when evaluation is not performed, thereby reducing the load.

The server 120 can also determine the timing of learning in accordance with the determined evaluation cycle. In this way, for example, if it is determined that only monthly processing will be performed, learning can be prevented from being performed daily or weekly, even though evaluation is performed monthly, thereby reducing the load.

Furthermore, in the initial stage, daily processing, which is the shortest evaluation period, is set to be performed. Here, for example, if weekly or monthly processing is set to be performed in the initial stage, if it is decided to perform daily processing after the initial stage, it may be impossible to create data for the daily processing from data for the weekly or monthly processing. However, in this embodiment, since daily processing is set to be performed in the initial stage as described above, even if it is decided to perform weekly or monthly processing after the initial stage, data to be used for weekly or monthly processing can be easily created.

In the evaluation period determination process according to this embodiment, in order to determine whether to perform evaluation for at least one of the multiple evaluation periods (monthly processing, which is the longest evaluation period), calculation of the average number of times the target appliance is used per month is omitted in step S401. However, it is also possible to determine, for example, whether the updated monthly average value is above a minimum standard, and then determine whether to perform monthly processing based on the determination result.

In addition, in this embodiment, 1 is used as the minimum standard, but this is not limited to this and any value can be used. Furthermore, the minimum standards for the daily average value and the weekly average value do not have to use the same value, but can use different values.

As described above, in the lifestyle pattern determination system 100 according to this embodiment,
a power sensor 110 (behavior detection unit) that can continuously detect the usage status of equipment installed in a house 1 (building) used by a subject P1, and thereby detect the behavior of the subject P1;
a server 120 (data creation unit) that creates learned behavior data related to the behavior of the subject P1 by learning measurement data (detection results) from the power sensor 110 (behavior detection unit);
a server 120 (condition setting unit) that sets conditions for the creation of the learning behavior data by the server 120 (data creation unit) so that the learning behavior data does not deviate from the actual behavior pattern of the subject P1;
It is equipped with the following.

In addition, a behavior detection step of detecting the behavior of the subject P1 by continuously detecting the usage status of equipment installed in the house 1 (building) used by the subject P1;
a data creation process of creating learned behavior data related to the behavior of the subject P1 by learning the measurement data (detection results) in the behavior detection process;
a condition setting step of setting conditions for creating the learning behavior data in the data creation step so that the learning behavior data does not deviate from the actual behavior pattern of the subject P1;
It is equipped with the following.

This configuration makes it possible to create learning behavior data that is in line with the actual behavior patterns of subject P1.

Furthermore, in the lifestyle pattern determination system 100,
The server 120 (condition setting unit)
One of the conditions is:
When the power sensor 110 (behavior detection unit) detects the use of the equipment for the first time, the server 120 (data creation unit) starts learning (step S105).

This configuration allows learning to begin at the appropriate time for each target appliance, making it possible to create learning behavior data that is in line with the actual behavior patterns of target person P1.

Furthermore, in the lifestyle pattern determination system 100,
The server 120 (data creation unit)
Based on the measurement data (detection results) obtained by the power sensor 110 (behavior detection unit), it is possible to create actual behavior data relating to the behavior of the subject P1, and to evaluate a comparison result between the actual behavior data and the learned behavior data (step S206).
The server 120 (condition setting unit)
One of the conditions is:
If the comparison result is evaluated to be outside the predetermined tolerance range, a learning necessity determination process is executed to determine whether or not the corresponding detection result (unlearned data), which is the detection result corresponding to the execution behavior data, needs to be learned (steps S209 and S301).

This configuration makes it possible to prevent the learning of measurement data that should not be learned, and to create learned behavioral data that is in line with the actual behavioral patterns of subject P1.

Furthermore, in the lifestyle pattern determination system 100,
The server 120 (condition setting unit)
In the learning necessity determination process, a request for feedback regarding the need to learn the corresponding detection result is sent to a predetermined notification unit (for example, a device portable by family member P2), and a decision is made as to whether the corresponding detection result needs to be learned or not based on the received feedback (steps S302, S303, and S307).

This configuration allows measurement data to be learned based on feedback, and learning behavioral data can be created that matches the actual behavioral patterns of subject P1.

Furthermore, in the lifestyle pattern determination system 100,
The server 120 (condition setting unit)
In the learning necessity determination process, if the corresponding detection result indicates that learning is not necessary, a determination is made as to whether the learning necessity determination process can be executed the next time a detection result with the same content as the corresponding detection result is detected (step S305).

This configuration makes it possible to prevent repeated requests for feedback on measurement data that does not require confirmation of whether family member P2 needs to study, for example.

Furthermore, in the lifestyle pattern determination system 100,
The server 120 (data creation unit)
The comparison result of the executive behavior data with the learning behavior data can be evaluated for at least two or more periods (daily (day), weekly (week), and monthly (monthly)) having different periods from each other;
An average value relating to the use of the target home appliance (equipment) is calculated for each period, and an evaluation is performed for each period among the multiple periods in which the calculated average value satisfies a predetermined threshold value.

This configuration makes it possible to determine the evaluation period (when to perform the evaluation), thereby improving the accuracy of the evaluation results.

Furthermore, in the lifestyle pattern determination system 100,
The facility includes a plurality of home appliances;
The server 120 (condition setting unit) sets the conditions for each of the plurality of home appliances.

This configuration makes it possible to create more accurate learning behavior data that is in line with the actual behavioral patterns of subject P1.

The lifestyle pattern determination system 100 according to this embodiment is an embodiment of a learning system.
The house 1 according to this embodiment is one embodiment of a building according to the present invention.
The power sensor 110 according to this embodiment is an embodiment of a behavior detection unit according to the present invention.
The server 120 according to this embodiment is an embodiment of a data creation unit and a condition setting unit according to the present invention.

For example, the lifestyle pattern determination system 100 according to this embodiment has been described with an example in which an elderly person is the subject P1, but the present invention is not limited to elderly people, and various people can be the subject P1.

In addition, in this embodiment, the power sensor 110 (behavior detection unit) is installed in the residence 1, but the present invention is not limited to this and can be installed in various buildings used by the subject P1. In other words, it is possible to detect abnormalities in cognitive function not only in the residence 1 but also in various other buildings, facilities, etc. Furthermore, the behavior detection unit is not limited to a power sensor, and various sensors capable of detecting human behavior can be used. Specifically, the behavior detection unit can be a human presence sensor or a recognition means capable of recognizing human behavior from captured images.

Furthermore, in this embodiment, an example has been shown in which server 120 (such as a virtual server installed on the cloud) performs various processes, but the present invention is not limited to this, and the entity that performs the various processes can be changed as desired. For example, the processes could also be performed by a home server, a personal computer, a portable terminal, etc. installed in residence 1.

In addition, in this embodiment, an example has been shown in which if an abnormality in the cognitive function of subject P1 is detected, the terminal 130 is notified of this. However, it is also possible for family member P2 or the like to confirm the abnormality and, if they determine that there is no problem (that there is no abnormality), to communicate this to the server 120 using the terminal 130 or the like. In this way, by having family member P2 or the like confirm the presence or absence of an abnormality and provide feedback to the server 120, the server 120 can learn the behavior of subject P1 more accurately. This allows for more accurate detection of abnormalities in cognitive function.

In addition, the period covered by each process exemplified in this embodiment (daily process, weekly process, monthly process, weekly result notification, and monthly result notification) can be changed as desired.

In addition, while this embodiment is intended to detect the presence or absence of abnormalities in cognitive function, it is not limited to this and can also detect the presence or absence of abnormalities in lifestyle patterns. For example, when installed in the home of someone who spends time away from their family, such as someone working away from home or a student living alone, it can detect the presence or absence of abnormalities in the lifestyle patterns of distant family members.

100 Lifestyle pattern determination system 110 Power sensor 120 Server 130 Terminal

Claims (6)

a behavior detection unit that can continuously detect the usage status of equipment installed in a building used by a subject person, thereby detecting the behavior of the subject person;
a data creation unit that creates learned behavior data related to the behavior of the subject by learning the detection result by the behavior detection unit;
a condition setting unit that sets conditions for the creation of the learning behavior data by the data creation unit so that the learning behavior data does not deviate from the actual behavior pattern of the subject;
Equipped with
The data creation unit
The action detection unit is capable of generating action data relating to the action of the subject based on the detection result, and is capable of evaluating a comparison result of the action data with the learned action data;
a comparison result between the active behavior data and the learned behavior data can be evaluated for at least two or more periods having different durations;
Calculating an average value for the use of the equipment in each period, and performing an evaluation for each period among the plurality of periods in which the calculated average value satisfies a predetermined threshold value.
Learning system. The condition setting unit
One of the conditions is:
when the behavior detection unit detects use of the facility for the first time, the data creation unit starts learning.
The learning system of claim 1 . The condition setting unit
One of the conditions is:
If the comparison result is evaluated to be outside a predetermined allowable range, a learning necessity determination process is executed to determine whether learning is necessary for the corresponding detection result, which is the detection result corresponding to the executive movement data.
The learning system according to claim 1 or 2. The condition setting unit
In the learning necessity determination process, a request for feedback regarding the learning necessity of the correspondence detection result is notified to a predetermined notification unit, and the learning necessity of the correspondence detection result is determined according to the obtained feedback.
The learning system according to claim 3 . The condition setting unit
In the learning necessity determination process, if the correspondence detection result indicates that learning is not necessary, a determination is made as to whether the learning necessity determination process should be executed when a detection result having the same content as the correspondence detection result is detected next time.
The learning system according to claim 3 or 4. The facility includes a plurality of home appliances;
the condition setting unit sets the condition for each of the plurality of home appliances.
A learning system according to any one of claims 1 to 5.