JP6489536B2 - Watch system - Google Patents

Description

  The present invention relates to a monitoring system for detecting daily behavior of a person to be watched such as an elderly person or a single person, detecting an abnormality of the person to be watched at an early stage, and watching the person to be watched safely.

  In order to realize a community where everyone can continue to live with peace of mind, it is an urgent task to build a watchover network for the elderly. The most important point in watch-over activities is to quickly notice any changes in the target, and to connect to specialized institutions and respond appropriately. However, it is difficult to quickly notice the subject's incidents only by external observations (the entrance lights are kept on, newspapers are collected in newspaper trays, face-to-face meetings are stopped, etc.) It is in the situation. Therefore, it is necessary to construct a mechanism that can observe the behavior of the house during the day and night with high accuracy while reducing the psychological barrier.

  Conventionally, there has been proposed a method of monitoring a single person by monitoring the single person's behavior and judging whether the behavior is a daily behavior or an extraordinary behavior. For example, a plurality of living rooms or living spaces are divided into a plurality of times, and the staying time of each section for each time zone is detected to determine the behavior of a solitary person (Patent Document 1). When it is determined that there is an abnormality by comparing the set watching condition and the behavior of a single person, it is set to report to the mobile phone with the TV phone function on the watching side (Patent Document 2), and the abnormality of the single person by the monitoring sensor When a condition is detected, an automatic notification system is used so that an abnormal notification communication can be made from a cell phone or a landline phone of a single person to a preset notification destination (Patent Document 3). For example, there is a method (Patent Document 4) for detecting an abnormality of a single person based on the duration of no operation of the home appliance.

JP 2005-157556 A JP 2008-210093 A JP 2012-89106 A JP 2013-178852 A

In order to detect a change in the person being watched for early, it is necessary to monitor the normal life pattern of the person being watched over. However, in order to detect noninvasively while ensuring the privacy of the subject, it is necessary to use sensors that can accurately determine the behavior of the subject. In addition, in order to grasp the life pattern of the subject throughout the day, it is necessary to detect the life pattern throughout the day and at night.
The present invention accurately detects the subject's life pattern based on the results of examining analysis items suitable for the purpose of non-invasively grasping the subject's life pattern while ensuring the subject's privacy. An object of the present invention is to provide a watching system that enables the target person to be watched accurately.

The monitoring system according to the present invention includes a sensor that detects behavior / health information related to the life pattern of the person being watched over, a sensor agent that monitors the sensor and collects behavior / health information obtained from the sensor, behavioral and health information collected by the sensor agent is transmitted to a watch system and a server detects the accident of watching target person based on the collected behavior and health information, as the sensor, the moving body sensors includes a remote controller sensor, a key ring fitted with a RFID, a bed sensor for detecting a change in posture on the subject's bed, and a pillow sensor for detecting the movement of the upper from the breathing or neck of the subject, the The sensor agent includes a data receiving means for receiving data from the sensor, and a sleeping / standby provided separately from the data receiving means. A plurality of data processing means provided corresponding to observation items set in advance, such as different floors, different whether they are at home or going out, and the data receiving means uses common data for sensors having a common data receiving method. In addition to receiving the data in the reception block, the data receiving block is provided with a distribution processing unit that distributes the data to each of the data processing units, and the amount of data required for each data processing is provided in the preceding stage of each data processing unit. As described above, message queue means for temporarily storing data output from the data distribution means is provided, and data relating to behavior / health information from the pillow sensor and the bed sensor is stored in real time in the server. Message queue length set so that no data is sent with In addition to providing queuing means, the urgent information is characterized by providing message queuing means in which the length of the message queue is set to a minimum data amount that enables real-time detection .
The behavior / health information defined in the present invention means at least one of behavior information and health information, and the information collected by the sensor / agent includes the behavior information or health information depending on the application. On the other hand, it becomes both action information and health information.
The data processed by the data processing means is further input to the data processing means at the next stage via the message queue means.
The distribution processing means can distribute data from one sensor received by the data reception block to a plurality of the data processing means.
The distribution processing means can distribute data from the plurality of sensors received by the data reception block to one data processing means.

The motion sensor used in the watching system according to the present invention detects the movement of the subject in the house, and whether the subject is at home or not based on the behavior / health information from the motion sensor. It is determined whether or not.
The remote control sensor is a sensor that detects infrared rays, and is used to detect that the subject has operated home appliances using the infrared remote control. Even when the subject is sitting at home and there is no movement, it can be determined from the behavior / health information of the remote control sensor whether the subject is in a healthy state.
An RFID-attached key ring is an RFID tag attached to a key used to lock the entrance and entrance of a room. Behavior / health information of an RFID-attached key bundle is based on whether the subject is at home or goes out. It can be used as information for determining whether or not

The sensor agent monitors the behavior and health information from said sensor, and a base station comprising means for collecting behavioral and health information, tablet device the Base Station action and health information collected in is transmitted With.
Embedded software is installed in the base station as a means for monitoring sensors and processing behavior / health information transmitted from the sensors. The collected behavior / health information is transmitted to the tablet terminal and transmitted from the tablet terminal to the server.

In addition to the sensors described above, the sensor can be provided with a physical switch that is operated spontaneously when the subject performs a specific action. The physical switch is used when performing a specific action such as pressing the switch when the subject goes out and pressing the switch when returning from the outside. The behavior / health information obtained from the operation of the subject pushing the physical switch can be used in combination with the behavior / health information collected from other sensors to determine whether the subject is healthy. Effective in terms.
In addition, as another sensor, a remote sensor that transmits behavior / health information about the subject using home appliances to the sensor / agent can be provided, so that the state of the subject can be further accurately determined. it can.

In addition, the server analyzes the life pattern of the target person's normal life based on the collected behavior / health information, and when the server detects a deviation from the normal life pattern of the target person, By providing the means for transmitting the alarm report, it is possible to accurately watch the change of the subject and notify the person in charge, thereby making it possible to make the watch of the subject more reliable.
In addition, when the server detects an operation in which the pillow sensor is artificially repeatedly shaken, the server is provided with means for sending an emergency call to the person in charge of watching, thereby notifying the person to be watched of the emergency situation. It becomes possible to perform an appropriate treatment. The pillow sensor is equipped with a triaxial acceleration sensor, and an emergency call can be made by shaking the pillow sensor when the subject has a sudden illness at bedtime.
In addition, the server includes means for analyzing the urgency of the anomaly that deviates from the normal life pattern of the target person, and means for transmitting an alarm report to the person in charge of watching as necessary based on the analysis result. Thus, the target person can be supported accurately according to the urgency of the target person.

In addition, the sensor agent does not send real-time information to the server for the behavior / health information from the pillow sensor and the bed sensor, and performs data processing for bedtime / wake-up determination by the data processing means. transmitted to the other said moving body sensor, the remote sensor, the key tag sensor for detecting the RFID mounted on the key ring is characterized that you send it to the server from other of said data processing means.

  According to the watching system according to the present invention, the life pattern of the watching target person can be analyzed and grasped non-invasively while protecting the privacy of the target person. Abnormality can be reliably detected, and the detection result can be suitably used for watching the subject.

It is explanatory drawing which shows schematic structure of the watching system which concerns on this invention. It is the external appearance photograph of the testing machine of a sensor agent. It is explanatory drawing which shows the system configuration | structure of the sensor agent which collects the action and health information from various sensors. It is a block diagram of a functional module with which a sensor agent is provided. It is a graph which shows the actual collection example of action and health information. It is a figure which shows the data processing in a sensor agent. It is a figure which shows the data processing in a sensor agent. It is a graph which shows the experimental result by a general purpose data processing method.

The most important point in the monitoring activities of the target person such as a single person is to notice the change of the target person as soon as possible. For this purpose, it is necessary to determine the threshold for rushing by combining the external observation of the subject and the observation of the in-house situation.
Information about the daily behavior of the subject can be collected and determined with high accuracy by increasing the number of observable points and observation items. However, when the number of observation points and observation items is increased, there is a risk that the psychological barrier of the subject increases, monitoring stress, and the risk of invading the subject's living body increases. In consideration of such restrictions, in the monitoring system according to the present invention, the sensor for detecting the life pattern of the subject is assumed to be completely non-invasive, and the observation items and the rushing action items of the subject are as follows: Set as follows.

The observation item of the subject is an item to be observed in order to detect the life pattern of the observer. The following items are set as observation items.
(A1) Sleeping / Waking up
(A2) At home or out
(A3) Operation and usage of TVs and air conditioners
(A4) Frequency of moving in the home
(A5) Confirm that you are not bedridden
(A6) Presence / absence of emergency call Emergency call is to make it possible for the subject person to watch the person himself / herself in an abnormal state and to report to the side.

In the rushing action item, the threshold value of whether the current state is within the normal range or greatly deviated (abnormal state) is compared with the normal life pattern of the person being watched over, and the abnormal state is detected. Need to be able to report abnormal conditions. The following items are set as rush action items.
(B1) Data collection and learning during normal times
(B2) An alert can be issued
(B3) Automatic notification to rush-in responders (relatives, local welfare officers, administrative staff, managers of housing for the elderly)
(B4) Immediate alert notification at the time of emergency call from the subject
(B5) Priority of rushing action
(B6) Utilization of collected data in an anonymized state as big data

In the watching system according to the present invention, in order to detect the behavior of the subject in the house with high accuracy throughout the day and night, a combined sensor specialized for living behavior in the daytime and at night is used in combination.
As sensors for detecting daytime living behavior, moving body sensors, remote control sensors, key tag sensors, physical sensors, etc. are used. As sensors for detecting nighttime (sleeping) living behavior, existing bed sensors and pillow sensors are used. To do. This bed sensor and pillow sensor were developed for the purpose of watching a single person at bedtime (Nobuhiro Shimoi, Hirokazu Mazokoro: “Study on a bed monitoring system using a three-axis acceleration sensor and a piezo load sensor”) , Transactions of the Society of Instrument and Control Engineers, 49 (12), 1092-1100 (2013)).
The bed sensor is provided with a load sensor on the upper surface of the bed, and detects the state of the subject at bedtime (such as turning over) and the change in posture when waking up. The pillow sensor has a three-axis acceleration sensor installed inside the pillow to detect the subject's sleeping situation in a more complex manner.

As a method for detecting a subject's behavior using a bed sensor and a pillow sensor, the present invention avoids collecting real-time behavior at bedtime from the viewpoint of psychological barrier and privacy protection of the subject, Considering the situation that requires, the following settings are made.
(1) Observed data of bed sensor / pillow sensor is not transmitted / transmitted in real time
(2) Receive and pre-process observation data with sensors and agents installed in the home, and determine the situation during sleep, before waking up, and after getting out of bed
(3) If you stay on the bed longer than normal, it is determined that you are bedridden
(4) When an emergency such as sudden illness is required at bedtime, the subject himself / herself shakes the pillow vigorously (referred to as “pillow shake”). A combination of a combination of sensors that specializes in daily living behavior and a sensor that specializes in nighttime living behavior can be used to accurately analyze the lifestyle patterns of the subject during normal times, resulting in changes in the subject. It becomes possible to judge exactly.
In addition to the pillow sensor and the bed sensor, the pillow sensor and bed sensor set the acceleration sensor on the underside of the bed bedding and the bed frame, in addition to information related to the sleep and waking up of the person being watched over. It is also possible to obtain information related to the health of the baby (vital information such as heart rate and respiration). Therefore, it is possible to detect the abnormality of the subject based on the behavior information related to the behavior of the subject and the health information of the subject.
The priority of (B5) rushing action in the rushing action item mentioned above can be set based on the target person's health information or health information and action information, and an immediate alert based on the urgency of the analysis result It can also be set to issue a report.

(Schematic configuration of the monitoring system)
FIG. 1 shows a schematic configuration of a watching system according to the present invention.
The monitoring system according to the present invention uses various sensors to collect information (behavior / health information) related to the life pattern of the person being watched over (behavior / health information). -Collecting health information, sending the collected behavior / health information to the server to learn the target person's normal life pattern, and anomalies (anomalous state) that deviate significantly from the normal life pattern obtained by analysis In this case, the system quickly detects the incident and immediately reports it to the person in charge of watching over the network to enable rushing action.
In addition to the regular behavior patterns such as waking up, going to bed, and eating that the subject performs almost every day, the subject's lifestyle patterns include periodic events such as going out on specific days of the week, and irregular behavior patterns. Including.

Sensors for obtaining behavior / health information related to the lifestyle patterns of the subject are those that ensure the subject's privacy, reduce the psychological barriers of the subject, and obtain behavior / health information non-invasively. use.
Sensors that detect the target person's life pattern can detect the target person's life pattern from multiple viewpoints so that the number of installations can be minimized and the target person's life pattern can be accurately grasped. Is used.

As a sensor to obtain behavior / health information related to the daytime lifestyle patterns of the subject, RFID is used to obtain information on the subject's home mode and absence mode, including a key used to lock the entrance and entrance of the room. Key bundle with tag attached, moving body sensor that detects the movement of the subject in the home, remote control sensor that detects that the subject operated the home appliance with the infrared remote controller, and the subject uses home appliances such as refrigerators and washing machines A remote sensor that detects this situation and a physical switch that operates according to the intention of the subject.
In addition, the bed sensor and the pillow sensor described above are used as sensors for obtaining behavior / health information related to the lifestyle pattern of the subject at night (at bedtime). Since the pillow sensor is provided with a triaxial acceleration sensor, an emergency signal can be captured by shaking the pillow sensor when the subject has a sudden illness at bedtime.

(Configuration of sensor / agent)
FIG. 2 shows a sensor agent testing machine manufactured for testing.
The sensor agent 10 includes a base station 20 and a tablet terminal 30 attached to the base station 20.
On the upper surface of the base station 20, a concave key bundle box 21 is provided for placing a key bundle 40 to which an RFID tag is attached. A tag sensor (not shown) that detects an RFID tag attached to the key bundle 40 is provided inside the base station 20 below the key bundle box 21.

In addition, a moving body sensor 41 and a remote control sensor 42 are provided at the front part of the casing of the base station 20, and a pair of push switches 22a and 22b are provided on the upper surface of the casing as physical switches operated by the person being watched over. Provided. In addition, an LED 23 is provided on the upper surface of the housing to indicate that the sensor agent is in operation.
Although not provided in the sensor / agent shown in FIG. 2, the base station 20 includes behavior / health information from the bed sensor and pillow sensor, and behavior / health information about opening / closing of the refrigerator and usage of household appliances. It is possible to provide means for detecting.

FIG. 3 shows a system configuration of the sensor agent 10 that collects behavior / health information from various sensors.
The wireless signal from the RFID tag attached to the key bundle 40 is detected by a non-contact sensor installed in the base station 20, the infrared signal when the remote controller is operated is detected by the remote controller sensor 42, and the movement of the subject is a moving body sensor. 41 and is collected by the sensor agent 10.
Remote sensors are attached to the bed sensor and the pillow sensor, and observation data from these sensors is transmitted to the sensor agent 10 via a wireless network and collected.
In addition, a remote sensor that detects opening / closing of the refrigerator is provided as a sensor that detects use of the refrigerator, and the opening / closing operation of the refrigerator is configured to be transmitted to the sensor agent 10 via a wireless network. For household appliances such as electric heaters, electric fans, washing machines, etc., household appliances are installed by attaching a wattmeter with a remote sensor to an outlet and transmitting the measured data of the wattmeter to the sensor agent 10 via a wireless network. Collect behavioral and health information on the usage status of
Behavior / health information related to the subject's voluntary movements such as operating a refrigerator and household appliances is information indicating that the subject is in a normal state.

(Behavior / Health Information)
The key ring 40 to which the RFID tag is attached serves as information for determining whether the subject is at home or going out, that is, information for determining whether the subject is in the home mode or the absence mode.
When the target person is at home, it is customary to place the key ring 40 attached with the RFID tag in the key ring box 21 of the base station 20, so that the target person's home mode and absence mode can be accurately managed together with the key ring collective management. Can be determined.
However, the key ring 40 attached with the RFID tag may be placed in a place other than the base station 20 in the home. For this reason, means for wirelessly detecting the RFID tag attached to the key ring 40 can be attached to the base station 20 to determine whether or not the key ring 40 exists in the home.
Since the RFID tag attached to the key ring 40 has an encryption function, there is an advantage that it can be used effectively for identification of the subject and is inexpensive.

By installing the moving body sensor 41 in the base station 20, it is possible to detect the behavior of the subject approaching or moving away from the sensor agent 10, and it is possible to obtain behavior / health information about the motion of the subject in the house. The sensor agent 10 is installed in a room often used by the subject during the daytime, such as a living room in the house, and the behavior of the subject is collected as behavior / health information.
Since the moving body sensor 41 may detect a dog or cat kept indoors in addition to the subject, the motion sensor 41 determines the motion of the subject together with behavior / health information obtained from other sensors.

  The remote control sensor 42 is for collecting operations using an infrared remote controller such as a television or an air conditioner as behavior / health information. When the subject is sitting in the living room and watching TV, the motion sensor 41 cannot detect the subject's movement. The remote control sensor 42 detects that the subject's remote control has been operated even when the subject's movement cannot be detected, so that the subject is at home and can operate a television or an air conditioner. It becomes action / health information in the state.

Since the remote control sensor 42 is a sensor for detecting infrared rays from the remote controller, the sensor agent 10 to which the moving body sensor 41 is attached is preferably installed in a place where the infrared rays from the remote control are easily received (such as near the television).
In the sensor agent 10 shown in FIG. 2, the moving body sensor 41 is mounted on the casing of the base station 20, but when there are a plurality of living rooms, the moving body sensor is installed as a remote sensor in the main living room or all the rooms. The behavior / health information may be transmitted to the sensor / agent 10 via the network.

The bed sensor and the pillow sensor can be used as behavior / health information when the subject is sleeping. The bed sensor serves as behavior / health information about body movements such as the subject turning over on the bed, and the pillow sensor serves as behavior / health information indicating the subject's breathing and small movements above the neck. The triaxial acceleration sensor installed in the pillow as a pillow sensor can detect even a small movement of the subject. Moreover, by attaching an acceleration sensor to the pillow sensor and the bed sensor, vital information (heartbeat, respiration, etc.) of the subject can be acquired.
Thus, the bed sensor and the pillow sensor are used as behavior / health information indicating the behavior and health of the subject at night (sleeping state).
The pillow sensor can also be used as an emergency call to notify a change by shaking the pillow when the subject suddenly develops a disease or the like. A trigger signal generated by shaking the pillow sensor is collected by the sensor agent 10, and when it is detected that the pillow is being shaken, it is immediately notified to the person in charge as an emergency signal. In addition, when the urgency of support is detected for the health condition of the subject based on the data acquired from the pillow sensor and the bed sensor, an emergency signal is notified to the person in charge of watching.

The behavior / health information collected by the sensor / agent 10 via the moving body sensor 41, the remote control sensor 42, the remote sensor, and the like described above is information collected regardless of the intention of the subject. On the other hand, the push switches 22a and 22b provided in the housing of the sensor / agent 10 are provided on the assumption that the subject person operates voluntarily.
The push switches 22a and 22b can be used for various purposes. For example, the push switch 22a is pushed when the subject goes out, and the push switch 22b is pushed when the subject comes back out. The behavior / health information for determining whether the person is at home or going out can be transmitted. Further, it is possible to use the push switch 22a when the user goes to bed and push the push switch 22b when the user gets up.

In addition, since the push switches 22a and 22b are provided in the casing of the base station 20, when the subject pushes the push switches 22a and 22b, the subject approaches the sensor / agent 10 by the moving body sensor 41. Therefore, it is possible to simultaneously detect an action such as leaving, and to confirm with high accuracy that the target person has taken the action. The motion sensor 41 alone may erroneously detect the movement of a dog or cat acting indoors as the movement of the target person. However, by using the push switches 22a and 22b together, such erroneous detection is prevented. can do.
Note that two push switches may be set as a set as in the above example, or each push switch can be used as a switch for identifying each different state. Further, the number of push switches is not limited to two, and physical switches (operation switches) other than push switches can be used.

  The behavior / health information obtained by the subject person's voluntary operation described above can be obtained as information with a high degree of accuracy regarding the subject's behavior compared to the moving body sensor 41 and the remote control sensor 42. The method of using the behavior / health information obtained by the motion sensor 41, the remote control sensor 42, or the sensor for detecting the operation of the home appliance in combination with the behavior / health information by the subject's spontaneous operation is based on the life pattern of the subject. Can be analyzed with higher accuracy, and it is effective as a method for more reliably determining an anomaly of the subject.

(Example of sensor / agent design)
FIG. 4 shows the configuration of each functional module (HW / SW) provided in the sensor agent 10.
The sensor agent 10 includes peripheral hardware such as push switches 22a and 22b and a moving body sensor 41, and also includes an embedded microprocessor 50 and a Bluetooth (registered trademark) connection interface 52. In the experimental example described above, Arduino: An Arduino: An Arduino multi-purpose shield with smartphone was used as the embedded microprocessor 50, and 1Sheeld: An Arduino multi-purpose shield with smartphone as the Bluetooth (registered trademark) connection interface.
As the tablet terminal 30, an Android (registered trademark) terminal having a screen size of 7 inches was used. The Android (registered trademark) terminal has a WiFi and Bluetooth (registered trademark) communication function, and behavior / health information from each sensor is transmitted to the base station 20 via Bluetooth (registered trademark) connection. The Android (registered trademark) terminal introduces a multipurpose connection application 1Sheeld App, and these applications are designed and implemented on the Text-to-Speech speech synthesis application N2TTS.

The actual design for the hardware is as follows.
Regarding the sensors, push switches 22a and 22b, a moving body sensor 41, a remote control sensor 42, a tag module for detecting RFID attached to the key bundle 40, and an LED 23 for operation confirmation are mounted on a mounting board. These are connected to the embedded microprocessor 50 via the SPI / PIO interface.
The behavior / health information is processed by the processor firmware (described later) and then transferred to the Bluetooth (registered trademark) connection interface 52 via the UART serial. The behavior / health information of the transferred sensor is transferred to the Android (registered trademark) terminal of the Bluetooth (registered trademark) pairing partner. The Android (registered trademark) terminal is designed to be sent to the network via the HTTP / TCP / IP stack.

The software was designed and implemented as follows.
The firmware on the embedded microprocessor performs edge triggering by high / low level detection from various sensors, initialization of the RFID tag module, and tag detection / departure determination.
The two push switches transmit “manual absence mode” and “manual home mode” on the left and right respectively. The moving body sensor performs a cumulative count of approaching / separating, and transmits the count value at the time of sensor separation trigger.
The RFID tag module detects the approaching and leaving of the Mifare tag attached to the key (bundle), and after each state continues for a certain period of time, it sends “automatic home mode” and “automatic absence mode” respectively.
The data received by the remote control sensor is transmitted as a result of encoding the manufacturer code and the data part of 32 bits.
The operation confirmation LED blinks when the push switch is pressed, when approaching is detected by the motion sensor, and when approaching is detected by the RFID tag.

After receiving behavior / health information sent from Arduino firmware by multi-purpose connection application 1Sheeld App installed on Android (registered trademark) terminal, function by base station (Logger, Clock, Terminal, HTTP, TTS) Execute the process.
The behavior / health information is set to be sent over the network via the WiFi line via the HTTP / TCP / IP stack.
When switching to home / absence mode manually / automatically, the text-to-speech speech synthesis application N2TTS is used to provide feedback to the target audience.

(Behavior / health information collection example)
In order to evaluate the actual data collection capability of the test machine described above, a small-scale demonstration model room was constructed and installed in the model room for data collection experiments. The model room is equipped with a foldable cot, a TV that can be operated with an infrared remote controller, and various fixtures (refrigerator, electric kettle, desk, chair, telephone).
FIG. 5 shows an example of collecting behavior / health information in a model room. The collected behavior / health information is (1) to (4) below.
(1) Push switch push trigger
(2) Cumulative count value of motion sensor
(3) RFID tag approaching / leaving edge trigger
(4) Remote sensor reception data

Along with these POST data, the agent device ID, firmware version, and cumulative number of activations are transmitted with the URL argument, and these transmission data are inserted and stored in the database installed in the home server. The transmission protocol from the home sensor / agent to the home server is HTTP, and the case is POST Method.
For visualization by device, D3.js (Data-Driven Documents) was used as the visualization engine.

FIG. 5 shows the behavior / health information (1) to (4) described above in the time zone from 6 o'clock to 20 o'clock for 5 days for the home sensor agent with a specific ID.
It can be determined that the time zone on the line where the motion sensor is continuously detected is actually “at home”, such as when the subject is at home and watching TV or moving. Also, the location received by the remote control sensor indicates that the TV is being turned on / off or the channel is changed, and the target person can be judged to be “sound” in that sense. There are two days when the behavior / health information of the push switch is completely lost and the behavior / health information by the motion sensor is lost for a long time. This can be determined that the subject is “out of office / going out”.
This test result shows that the sensor / agent prototype can accurately collect behavior and health information.

  Since there are various variations in the normal life pattern of the watched person, it is necessary to analyze the life pattern of the target person over a period of time sufficient to understand the various variations. It is necessary to accumulate behavior / health information about. Actually, while accumulating behavioral and health information on the lifestyle patterns of the subject, we proceeded with the analysis of lifestyle patterns, and also examined the threshold values indicating the health status and deviation from the health status. It is necessary to improve the detection accuracy.

(Data processing method)
In the monitoring system according to the present invention, as an observation item of the subject,
(A1) Sleeping and waking up (A2) Staying at home and going out (A3) Operation and usage of TVs, air conditioners, etc. (A4) Frequency of moving in the home (A5) Confirming that they are not bedridden (A6 ) The presence or absence of an emergency call is set, and these items are observed using appropriate sensors such as pillow sensors and moving body sensors.
The output of the watching system includes an immediate alert notification at the time of an emergency call and a notification for rushing response to a deviation from a normal life pattern.
Of these outputs, emergency calls are detected by the sensor / agent due to their urgency, and then sent to the server in a short time to be output. Based on the transmitted behavior / health information, the server determines and outputs. Behavior / health information from sensor groups such as pillow sensors, bed sensors, moving body sensors, remote control sensors, key tag sensors, etc. are sequentially sent to the server as basic data. Judgment is made as to whether it is in a life pattern state or an anomaly state that deviates from the normal life pattern.

FIG. 6 shows an example of data processing in the sensor agent for data transmitted from the sensor group to the sensor agent.
As shown in FIG. 6, the sensor / agent receives behavior / health information from a pillow sensor, a bed sensor, a moving body sensor, a remote control sensor, and a key tag sensor and transmits them to the server. Of the behavior / health information received by the sensor / agent, alarm information is transmitted from the sensor / agent as emergency information for emergency information due to the swing of the pillow sensor (swing detection). On the other hand, information from sensors other than emergency information is individually transmitted from the sensor / agent to the server as basic data and sequentially stored.

Here, the transmission data transmitted from each sensor to the sensor agent includes data that can be used as meaningful data such as a moving body sensor, a remote control sensor, and a key tag sensor, a pillow sensor, and a bed sensor. Like a sensor, there are some data that cannot be used as data for determination unless simple data processing is performed because the amount of information is too large if it is raw data. In other words, the data transmitted from the moving body sensor or the remote control sensor is the trigger behavior / health information transmitted intermittently, and the pillow sensor or the bed sensor transmits several tens of data per second. Continuous behavior / health information.
As described above, the behavior / health information sent to the sensor / agent varies depending on the characteristics of the sensor used as the sensor and the role of each sensor. Are mixed.

In the sensor / agent shown in FIG. 6, the behavior / health information from the pillow sensor and the bed sensor is transmitted to the server by performing data processing for bedtime / wake-up determination in the sensor / agent, and other moving body sensors and remote controllers. The sensor and key tag sensor are transmitted to the server as they are.
As shown in FIG. 6, the behavior / health information transmitted from the sensor can be processed by setting a processing method for each sensor. However, in the monitoring system, the monitoring accuracy is improved. In addition, it is easily assumed that changes such as changes in data types, changes in sensors, addition of new sensors, and addition of new processes occur. In such a case, the method of constructing a system by changing the combination of data from the sensor and its processing for each sensor has a complicated system configuration, making various corrections and adding functions complicated. become.

Hereinafter, a framework for easily and versatilely performing data reception processing and data processing from a sensor even when there is a change in data type or sensor change will be described.
Generalizing the data processing requirements in the sensor agent is as follows.
(R1) Data from one sensor is used for multiple processes (R2) Data from multiple sensors is used for one process (R3) The amount of past data required for each process differs (R4) Sensors may be added (R5) Processes may be added In the monitoring system of the present invention, behavior / health information obtained from various sensors is used as a method for judging deviations from normal life patterns. It is characterized in that it is used in combination. In addition to accumulating behavior / health information received by sensors / agents, it aims to improve watching accuracy by combining behavior / health information.

FIG. 7 shows an example of data processing in the sensor agent that satisfies the processing requirements described above.
In order to satisfy the above processing requirements, first, data reception from the sensor and data processing operation are separated. By separating the data reception from the sensor and the data processing operation, it is possible to flexibly cope with the addition of the sensor and the addition of the data processing, and the requirements (R4) and (R5) can be satisfied.
Note that the data processing in FIG. 7 means data processing for observation items such as sleeping / wake-up, home / outing, movement between a plurality of motion sensors, and usage status of home appliances.

  The reception of data from the sensor is divided into data reception blocks (data reception means) that depend on the data reception method, and sensors having the same data reception method are received by the common data reception block. In the data reception block, a distribution process (distribution processing means) is performed to which data processing block (data processing means) the data transmitted from the sensor is distributed. This distribution process satisfies the requirement (R1) of using data from one sensor for a plurality of processes.

  The reason why the data reception block is provided depending on the data reception method is that the data communication method between the sensor and the sensor agent depends on the sensor. For example, when data communication between a sensor and a computer (sensor agent) is performed by serial communication such as I2C, a data reception process based on a serial communication processing program is required. Therefore, the data receiving process is implemented not as each sensor but as a data receiving block (data receiving means) depending on the data communication method with the sensor as shown in FIG.

As described above, since data reception and data processing are separate processes, it is necessary to exchange data between processes. For this reason, a message queue block (message queue means) having a required length for adjusting the data processing timing is provided in the preceding stage of the data processing block (data processing means).
This message queue block is provided as an independent process so that it can accept data from a plurality of data receiving blocks. By providing the message queue block, it is possible to satisfy the requirement (R2) of using data from a plurality of sensors.
The message queue block is provided by appropriately adjusting the length of the message queue so that necessary data can be stored in the data processing block. By adjusting the length of the message queue, even if the amount of past data required for each data processing is different, the requirement (R3) of appropriately using data from the sensor for data processing can be satisfied.

  As shown in FIG. 7, the message queue block can be configured to input the data output from the data reception block before the data processing block. A configuration may be adopted in which data is input to the message queue block together with sensor data. In other words, the message queue block is arranged in front of the data processing block to accumulate data necessary for the data processing, and is not limited to data from the sensor. Can be used as input.

In the data processing shown in FIG. 7, since the emergency call processing when it is detected that the pillow has been swung is required to be real-time, it is necessary to perform detection processing every time the most recent data that needs to be detected accumulates. . Therefore, the emergency call process can be accurately performed by setting the length of the message queue to the minimum necessary data amount that requires this detection. Accordingly, emergency call processing can be performed by the data processing method shown in FIG.
For data transmission to the server, the minimum necessary data accumulation can be realized by specifying the length of the message queue based on the communication interval considering the transmission overhead via the Internet. For emergency call processing, it is necessary to have both real-time performance and certainty of data delivery. Although data transmission via the Internet takes a relatively long time, both the real-time property and the prevention of overflow of the transmission buffer are realized by appropriately controlling the queue length by the message queue method.

(Data processing experiment)
An experiment was made to test whether the requirements (R1), (R2), and (R3) were satisfied by the data processing method described above.
RaspberryPi 3 Model B was used for the prototype computer system. In the experiment, behavioral and health information was collected using pillow sensors and bed sensors. About 10 points of data are transmitted from the used sensor at 50 Hz, and there is a margin in processing capacity.
Regarding the separation of processing processes, an OS having a multitasking function was installed in a data processing computer, and the simultaneous execution of multiple processes was managed by the OS function. Specifically, Linux (registered trademark) was used as a multitasking OS, and Raspbian Jessie, one of Linux (registered trademark) distributions for RaspberryPi, was used. Each process was developed as a separate program on this Linux (registered trademark), and parallel processing was realized by executing it simultaneously.

For the message queue, we used Redis, an open source software. Redis is a type of database server that runs in memory.
Although Redis is a database server, it has a function of holding a list of character strings. By handling this list in a first-in first-out manner, it can be used as a message queue. Further, since it is a database server, it supports data transmission / reception through inter-process communication or network communication, and can satisfy the functions of sensors and agents.

For data reception, data communication with the sensor was performed by ZigBee (registered trademark).
The pillow sensor used in the experiment is a 3-axis acceleration sensor for pillows, and the bed sensor is a 5-channel load sensor. The triaxial acceleration sensor transmits the acceleration of each axis every 20 ms. The load sensor also transmits data via Zigbee (registered trademark) every 20 ms.

For data processing, we implemented three types of processing: shaking detection, bedtime / wake-up determination, and transmission to the server.
Pillow swing detection was performed by detecting a specific frequency from the acceleration data using FFT, and was implemented using Perl. The message queue length is set as required for FFT.
For bedtime / wakeup determination, we implemented a process that detects vibrations from human movements from pillow and bed sensors, and determines whether or not you are on the bed. Since this process requires 1 second of data, a message queue capable of accumulating data of 2 seconds was prepared with a buffer margin. For this implementation, we used Python with a track record of creating motion detection programs.
Sending to the server is a process of sending the determination results in each process to the server via the Internet. Considering the clouding of services, perform processing based on HTTP clients with a proven track record in server connections. And implemented with Perl. In consideration of the overhead of data transmission by HTTP and the delay of data resin on the server side, the transmission interval is decided to transmit data collectively every 5 seconds. As a result, the message queue is also set to a length capable of storing data for 5 seconds.

Fig. 8 shows the result of creating a program that visualizes the results received on the server side and confirming the operation of the prototype system. The data collection results obtained in the experiment are shown as a graph. It is.
In FIG. 8, BED indicates the analysis result of the data from the bed sensor, and PIL indicates the analysis result of the data from the pillow sensor.
In the figure, part a (dark color) indicates that the sensor is operating but no motion is detected, part b (light color) indicates movement, and part c (intermediate dark color) detects oscillation. Each state is shown.
The experimental results shown in FIG. 8 indicate that data reception / data processing processes independently developed operate in cooperation with each other. That is, the processing method shown in FIG. 7 can be used for data processing from the sensor.
In addition, according to the results of data processing using the acquired behavior / health information, priority is given to rushing support such as “Danger”, “Slightly dangerous”, and “Need absolute confirmation” as rushing action items to the watching side It can also be configured to display on the reception screen on the person's side and notify the supporter of the urgency level.

10 Sensor agent 20 Base station 21 Key ring box 22a, 22b Push switch 23 LED
30 Tablet Terminal 40 Key Ring 41 Motion Sensor 42 Remote Control Sensor 50 Embedded Microprocessor 52 Bluetooth (Registered Trademark) Connection Interface

Claims (9)

A sensor that detects behavior and health information related to the life pattern of the person being watched over,
A sensor agent that monitors the sensor and collects behavior / health information obtained from the sensor;
A monitoring system comprising a server that transmits behavior / health information collected by the sensor / agent and detects a change in the person being watched based on the collected behavior / health information ,
As the sensor, a moving body sensor , a remote control sensor , a key bundle equipped with RFID, a bed sensor for detecting a change in posture on the subject's bed, and detecting the subject's breathing and movement from the neck With a pillow sensor,
The sensor agent is
Data receiving means for receiving data from the sensor;
A plurality of data processing means provided corresponding to observation items set in advance, such as sleeping / wake-up, home / going-out, provided separately from the data receiving means,
In the data receiving means, sensors having a common data receiving method are received by a common data receiving block, and further, a distribution processing means for distributing the data received by the data receiving block to the individual data processing means,
In front of each data processing means, a message queue means for temporarily storing data output from the data distribution means is provided so that the amount of data required for each data processing is obtained,
For data related to behavior / health information from the pillow sensor and the bed sensor, a message queue means is provided that sets the length of the message queue so that data is not transmitted to the server in real time. A monitoring system comprising a message queue means in which a queue length is set to a minimum data amount that enables real-time detection of a queue length . 2. The watching system according to claim 1, wherein the data processed by the data processing means is further input to the data processing means at the next stage via the message queue means. The monitoring system according to claim 1 or 2, wherein the distribution processing means can distribute data from one sensor received by the data reception block to a plurality of the data processing means. 4. The distribution processing unit can distribute data from the plurality of sensors received by the data reception block to a single data processing unit. The monitoring system described. The server includes means for analyzing a normal life pattern of the subject based on the collected behavior / health information;
The monitoring according to any one of claims 1 to 4, further comprising means for sending an alarm notification to a person in charge of watching when an abnormality deviating from a normal life pattern of the subject is detected. system.   The monitoring system according to claim 5, wherein the server includes means for sending an emergency call to a person in charge of watching when the operation of the pillow sensor being artificially repeated is detected. The server analyzes the urgency of anomalies that deviate from the normal life pattern of the subject; and
6. The watching system according to claim 5, further comprising means for transmitting an alarm report to a person in charge of watching based on the analysis result. The sensor agent comprises a base station comprising means for monitoring behavior / health information from the sensor and collecting behavior / health information;
Tracking System of any one of claims 1-7, characterized in that it comprises a tablet device the Base Station action and health information collected in is transmitted. The sensor agent does not send real-time information to the server for behavior / health information from the pillow sensor and the bed sensor, and transmits data to the server by performing data processing for bedtime / wake-up determination by the data processing means. and, other said moving body sensor, the remote sensor, the key tag sensor for detecting the RFID mounted on the key ring is claim 1 for the other of said data processing means, characterized that you send it to the server The monitoring system according to any one of 8 .

Images