WO2022091176A1

WO2022091176A1 - Behavior analysis system and behavior analysis method

Info

Publication number: WO2022091176A1
Application number: PCT/JP2020/040075
Authority: WO
Inventors: 健太郎佐野; 昭義大平; 佐知田中
Original assignee: 株式会社日立製作所
Priority date: 2020-10-26
Filing date: 2020-10-26
Publication date: 2022-05-05
Also published as: JPWO2022091176A1; CN115997233A

Abstract

The purpose of the present invention is to provide a behavior analysis system and a behavior analysis method which detect various behaviors of a subject and changes in these behaviors, and which improve monitoring accuracy. For this purpose, the present invention involves a behavior analysis system comprising an information acquisition unit, an analysis unit, and an output unit, said behavior analysis system being characterized in that: the information acquisition unit acquires data from at least one sensor that detects changes occurring due to behavior of a subject to be analyzed; the analysis unit classifies the data acquired by the information acquisition unit into one or more categories and assigns a unique identification symbol to at least one of the classification results; and the output unit outputs time series information about classification results to which the same identification symbol has been assigned.

Description

Behavior analysis system and behavior analysis method

The present invention relates to a behavioral analysis system and a behavioral analysis method.

In recent years, the birthrate is declining and the population is aging, and the need for watching over the elderly living alone is increasing. Patent Document 1 is a background technique in this technical field. According to this patent document, "In the sensor signal processing system, it is possible to quantitatively analyze the movement of the subject as an acceleration based on the body movement data acquired from the measuring device. The movement of the subject's body. The acceleration can be used, for example, for grasping the subject's condition such as the subject's behavior and health condition, and for identifying an individual. " According to this method, terminal care (end-of-life care), falls during walking, illness or injury requiring hospitalization, death, cognitive decline (cognitive impairment), wandering, etc. can be grasped. Furthermore, individual behaviors of the subject in daily life, such as getting up from sleeping (getting out of bed), excretion, and going to bed, can be detected.

Japanese Unexamined Patent Publication No. 2019-165071

The technique described in Patent Document 1 can detect individual behaviors of a subject in daily life based on body movement data. However, in order to watch over the elderly, it is important to pay attention not only to the occurrence of specific behaviors but also to changes in daily behaviors. Further, in the technique of the patent document, it is difficult to detect an event that the system designer does not anticipate or an action that cannot be specified only by the body motion data, so that there is room for improving the accuracy of watching.
An object of the present invention is to provide a behavior analysis system and a behavior analysis method that detect various behaviors of a subject and their changes and improve the accuracy of watching.

In order to solve the above problems, the present invention is a behavior analysis system including an information acquisition unit, an analysis unit, and an output unit, and the information acquisition unit is generated due to the behavior of an analysis target person. The data of at least one sensor that captures a change is acquired, the analysis unit classifies the data acquired by the information acquisition unit into one or more types, assigns a unique identification symbol to at least one of the classification results, and outputs the data. The unit is characterized by outputting time-series information of classification results to which the same identification symbol is assigned.

According to the present invention, it is possible to provide a behavior analysis system and a behavior analysis method that detect various behaviors of a subject and their changes and improve the accuracy of watching.

It is a block diagram which shows the 1st structure of the behavior analysis system which concerns on embodiment of this invention. It is a flowchart explaining the operation of a behavior analysis system. It is a figure explaining that the behavior analysis system of Example 1 classifies the behavior of the analysis subject using different kinds of sensors, and outputs the characteristic of the behavior. It is a block diagram which shows the 2nd structure of the behavior analysis system which concerns on the modification of this invention. It is a figure which shows the trader which has a behavior analysis system, and the related party.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings, but the present invention is not limited to the following embodiments, and various modifications and applications are also included in the technical concept of the present invention. Included in that range.

FIG. 1 is a block diagram showing the first configuration of the behavior analysis system of the present embodiment. The behavior analysis system 1 that senses and analyzes human behavior is composed of an information acquisition unit 11, an analysis unit 12, a storage unit 13, a display content determination unit 14, an output unit 15, and a conversion unit 16. It is built on a server connected to the network or on the cloud.

The information acquisition unit 11 has a function of acquiring data of one or more sensors that capture changes caused by the behavior of the analysis target person. Sensors are installed in spaces where humans live, such as in homes and facilities, and there are multiple acquisition methods depending on the sensor data collection method. For example, when collecting sensor data to a locally installed PC via a gateway, the information acquisition unit 11 acquires sensor data by accessing the PC via a local network or the Internet. As another example, when the sensor data is collected directly or via a gateway to a server that can be generally rented, the information acquisition unit 11 acquires the sensor data by accessing the server via the Internet. As yet another example, the sensor and the behavior analysis system 1 may directly transmit and receive information by wireless communication.

The types of sensors from which the information acquisition unit 11 can acquire data include, for example, a motion sensor 31, an illuminance sensor 32, a temperature / humidity sensor 33, a radio wave sensor 34, and an image sensor 35. Here, the radio wave sensor 34 is defined in the present specification as a general term for sensors that detect the existence and movement of an object using radio waves, such as a microwave sensor, a millimeter wave sensor, and a TOF sensor. The type of sensor is not limited to these, and may include, for example, a door open / close sensor (not shown), a pressure sensor, a microphone, a noise sensor, a vibration sensor, an acceleration sensor, an odor sensor, and the like. Further, even if the measurement contents are the same, sensors having different measurement principles, measurement intervals, and measurement value ranges may be included.

Further, the information acquisition unit 11 may pass through the conversion unit 16 when transmitting the sensor data to the analysis unit 12 and the storage unit 13. The conversion unit 16 has a function of unifying the format of the acquired sensor data. The conversion unit 16 prepares, for example, a unified format for describing time data and sensor data in 1-minute increments, and applies the format to a plurality of sensors having different measurement intervals. As a specific process of the conversion unit 16, for example, a sensor measuring every second may be calculated and converted into data in 1-minute increments by calculating values such as maximum, minimum, and average for 1 minute. Further, for example, the conversion unit 16 may calculate and estimate values such as an average of measured values and a linear approximation from the data for 4 minutes between the measurement times for the sensor that measures every 5 minutes.

The analysis unit 12 has a function of receiving and analyzing the sensor data transmitted by the information acquisition unit 11 directly or via the conversion unit 16. The analysis unit is composed of a classification unit 12a, a time series analysis unit 12b, and an index calculation unit 12c.

The classification unit 12a classifies one or more types of sensor data transmitted by the information acquisition unit into one or more types. Here, classification is to group data based on its characteristics, and it is not always necessary to assign human behaviors such as "sleep" and "going out", but if necessary, the allocation method is limited to human behaviors. You may. The classification method is not particularly limited. For example, even if one or more specific sensor data is focused on from a plurality of sensor data and the reaction is used to determine the classification of all the sensor data at the same time. good.

Alternatively, for example, using supervised machine learning, teacher data and learning data including cases to be classified in advance may be prepared, and a classifier capable of classifying data according to the number of types of teacher data may be trained and used. Here, an example of teacher data is, for example, human behavior such as "sleep" and "going out", and an example of learning data is, for example, sensor data in a time zone corresponding to the above-mentioned "sleep" and "going out". The types of machine learning are not limited as long as they learn using teacher data and acquire classification functions, such as decision tree methods including boosting, logistic regression, k-nearest neighbors, and support. A vector machine, a random forest and their ensemble, a fully connected layer of deep learning, a classifier constructed by utilizing CNN, RNN, etc. may be utilized.

Furthermore, for example, unsupervised learning may be used as a classification method, and data having similar feature quantities of the collected sensor data may be collected. The type of machine learning is not limited as long as it has a classification function without teacher data, for example, k-means method, k-means ++ method, x-means method, k-shape method, mixed Gaussian model, etc. Clustering method, a method using anomaly detection methods such as OneClass SVM, Elliptic Evolve, Isolation Forest, Local Outlier Factor, mixed Gaussian model, and a clustering method that combines them with dimension reduction methods such as t-sne and autoencoder. , The low-dimensional space of the hostile generation network represented by Advanced Autoencoders may be used, and a method of organizing data into a predetermined distribution may be utilized. In addition, the number of classifications is determined by using a method such as an arbitrary method, an elbow method, a silhouette method, etc. If there is something different than the value, it may be counted as a different classification. By using the classification method that does not require teacher data in this way, it is possible to capture and classify the characteristics of data even for events that the system designer does not anticipate.

The classification unit 12a can assign a unique identification symbol to the classification result. Here, the unique identification symbol includes human behavior such as "sleep" and "going out", as well as alphabets, numbers, images, feature amount vectors such as A and B exemplified in 12a and 12b of FIG. The type is not particularly limited as long as it can be identified by humans or machines that the classification result is different from other classification results. The method of assigning the identification symbol is not limited, and for example, the sensor data belonging to each classification result may be used as a clue. Specifically, identification symbols may be assigned based on the time range, the presence / absence of reaction of a specific sensor, the calculation results thereof, and the like, and as illustrated by 〇 and Δ in 12a of FIG. 1, t- It may be assigned based on the feature amount extracted by sne or the autoencoder and their similarity. Further, without explicitly giving an identification symbol, a unique identification symbol may be randomly assigned to a data group determined to form one classification as a result of utilizing machine learning.

The time-series analysis unit 12b performs time-series analysis based on the result of classifying the current data by the classification unit 12a or the data of the storage unit 13 that has accumulated them for a certain period of time. Here, the time series analysis focuses on one or more arbitrary values, analyzes how the values have changed with the change of time from the past to the present, and how the values will change in the future. It is a process to predict whether or not.

The time series analysis may target not only one specific classification result but also a plurality of types of classification results, their ratios, and calculation results such as the number of occurrences within a certain period and the difference in time. Further, as an example of the limit in which the time range is short, only the current data may be analyzed. As the time series analysis method, in addition to the k-shape method, classification using machine learning such as RNN and 1DCNN, regression and clustering, the above classification method may be applied in the time series direction. Further, for each classification result, the average, variance, covariance, standard deviation, etc. may be calculated for the number of occurrences, time, number of changes of the sensor during occurrence, and the like, and the time change thereof may be used. It should be noted that the time series analysis includes not only the analysis of past data but also the prediction of future behavior and sensor reaction based on the analysis.

The index calculation unit 12c uses the data of the storage unit 13 or the result of the time series analysis unit 12b to calculate an index that correlates with changes in human behavior. This index may be, for example, a numerical value or a graph of time-series changes such as the number of types of human behavior, time, ratio, and interval in a day. Specifically, as illustrated in 12c in FIG. 1, time-series changes such as the number of types of human behavior, time, ratio, and interval in a day are converted into three types of indexes, C, D, and E, and 3 For example, plot on the graph of the axis. In either case of digitization or graphing, a certain threshold value may be set and the magnitude of the measurement result with respect to the threshold value may be evaluated. This index may be calculated, for example, by quantifying the data of the storage unit 13 and the result of the time series analysis unit 12b on the three axes of meal, housework, and relaxation, and analyzing the changes in those time series. .. Specifically, for example, meal, housework, and relaxation are each evaluated on a 10-point scale, and the threshold value is 7. Regarding meals, 3 when the frequency of meals is detected once or more and less than 2 times in 24 hours, 6 when 2 times or more and less than 3 times are detected, and 9 when the frequency of meals is detected 3 times or more. May be determined to be given. Similarly, regarding household chores, 3 when 1 or more and less than 3 types of household chores are detected in 24 hours, and 5 or more in 6 or 24 hours when 3 or more and less than 5 types are detected in the time series analysis unit. It may be determined that 9 is given when it is detected. Similarly, regarding relaxation, when the time series analysis unit detects a relaxation time of 0.5 hours or more and less than 1 hour in 24 hours, 3 or 1 hour or more and less than 1.5 hours, 6, 1 It may be determined that 9 is given when it is detected for 5 hours or more. Alternatively, in the case of relaxation, it is important to balance the appropriate time and the number of relaxations, and if relaxation is detected for 0.5 hours or more and less than 1 hour continuously, it is 3, 1 hour or more and less than 1.5 hours continuously. When it is detected, 6 may be given, and when it is detected continuously for 1.5 hours or more, 1 may be given, and the final evaluation may be determined using the calculation result such as the product with the number of times and the sum. Similarly, for the relaxation time interval, 1 is given to the interval of 0.5 hour or more and less than 1 hour, 6 is given to the interval of 1 hour or more and less than 2 hours, and 3 is given to the interval of 2 hours or more. The final evaluation may be determined using the calculation result with the numbers given by the time and the number of relaxations. Furthermore, by analyzing the changes in time series of these quantified results, an index that correlates with the changes in human behavior may be calculated. To give an example of the analysis of changes in time series, the increase / decrease in the numerical values given to the above three axes is analyzed in time series, and it is predicted that the numerical values will fall below the threshold value.

The quantification method is not limited to these. For example, the three axes of meal, housework, and relaxation may be evaluated on a scale other than the 10 grades, or may be evaluated on a continuous value without any grade. Further, the evaluation period may be changed not only for 24 hours but also for one week and one month. Further, the value of meals may be calculated by using not only the number of meals but also the food, the time taken for meals, the people who ate together, the number of people, and the like. Housework may be calculated using not only the type of housework but also the time spent doing housework, exercises, bathing, and other household chores that are not classified as household chores. Relaxation may be calculated using not only the time, number of times, and interval, but also the number of people relaxed in the same space, the attributes of the other party, and so on. The threshold may be set differently for each of the three axes of meal, housework, and relaxation, and the total value, relative value, ratio, maximum value, minimum value, time-series change, and data of others on the three axes. The final index may be calculated by paying attention to the difference between the two. Furthermore, the types of the three axes may be determined other than meals, housework, and relaxation, and it is judged that the calculation of the values of the three axes is not only the data classified as human behavior but also mechanically characteristic. Classification results to which identification symbols such as alphabets, numbers, images, and feature amount vectors are assigned may be used. The index calculation unit 12c may calculate an index not only for changes in human behavior but also for other behaviors and symptoms and their precursors. For example, the walking speed may be calculated from the data related to the position information in the house, and the index correlated with the fall risk during walking may be calculated from the time-series change. In addition, an index that correlates with the risk of falling in the bathroom may be calculated from the data on the temperature difference between the rooms in the house and the change over time.

The storage unit 13 obtains the sensor data acquired by the information acquisition unit 11, the classification result of the classification unit 12a of the analysis unit 12, the analysis result of the time series analysis unit 12b, the calculation result of the index calculation unit 12c, and the conversion result of the conversion unit 16. accumulate. In addition to accumulating for each analysis target person, the terminal device 50 is tagged with attributes such as age, gender, origin, language, religion, hobbies and preferences of the analysis target person acquired by inputting using the input unit 17. You may. The users here are assumed to be individuals who want to watch over the analysis target such as elderly parents living alone, as well as insurance companies and day service companies.

The display content determination unit 14 acquires the result of the analysis unit 12, determines the content to be displayed to the user, and transmits it to the output unit 15. The method of determining the display content is selected from at least one candidate preset by the system designer according to the attributes of the individual or the trader who is the user of the behavior analysis system 1. Specifically, the system designer may preset the behavioral recommendation display, the daily life report display, and the like as options.

The behavioral recommendation display is selected, for example, when the user wants to watch over an elderly parent or the like, and can promote behavioral change to improve the health condition based on an index correlated with the behavioral change. .. The daily life report display is selected, for example, when the user is an insurer, and by applying this system to the insured's home, it is possible to periodically notify the insurer of indicators that correlate with changes in behavior. Conceivable. Further, for example, even when the user is a day service provider, the report display of daily life may be selected. By applying this system to the home of the day service subscriber, the daily life of the patient can be recorded in a schedule format. You can refer to it when you visit. The user (trader) who utilizes the behavior analysis system 1 is not particularly limited, and the courier, taxi company, retailer, distributor, etc. effectively provide the services and products of each company to the individual. You may utilize the information displayed from the system. Specific examples include reliable delivery based on home time, dispatching taxis based on outing habits, and determining recommended products based on lifestyle habits.

The display content determination unit 14 may determine the display content according to the input from the input unit 17 of the terminal device 50 possessed by the user. For example, regarding the result of the analysis unit 12, only the items specified by the terminal device 50 may be determined. May be displayed. Further, the display content determination unit 14 may score the previous output based on the feedback given from the terminal device 50 to the previous output, and leave only the one with a high score.

The output unit 15 outputs the result of the analysis unit 12 to the terminal device 50 based on the determination of the display content determination unit 14. The terminal device 50 as an output destination is a terminal such as a dedicated display or a user's smart device prepared by a company that provides a behavior analysis system, and has an input unit 17 and a display unit 18. Further, the transmission of the data displayed on the display unit 18 of the terminal device 50 can be realized by a wired connection, a wireless connection, an access right to the data on the server, or the like. The terminal device 50 may be configured as an element of the behavior analysis system 1.

FIG. 2 is a flowchart illustrating the operation of the behavior analysis system 1 of the present embodiment. The behavior analysis system 1 starts processing triggered by the operation of the operator of the system (S1). First, the information acquisition unit 11 acquires sensor data (S2). The installation environment of this sensor may be different for each user of the behavior analysis system 1. For example, when the user of the behavior analysis system 1 is an individual who wants to watch over his / her parents, the place where the sensor is installed is in the house of the person to be analyzed who needs to watch over. Further, for example, when the user of the behavior analysis system 1 is an insurer, the sensor is installed in the insured's house or facility. Further, for example, when the user of the behavior analysis system 1 is a day service provider, the sensor is installed at the subscriber's home.

After acquiring the sensor data, the information acquisition unit 11 confirms whether the unified data format is defined in the conversion unit 16 (S3), and if it is defined, the conversion unit 16 changes the data format accordingly (S4). .. In addition, this change method can be considered to change the file format, change the time interval of data, change significant figures, change the format, change the character string to numbers, etc. It may be defined as a unified format of data including. After that, the information acquisition unit 11 transmits the data to the analysis unit 12 and the storage unit 13 (S5).

The analysis unit 12 first analyzes the current data (S6) and transmits the result to the storage unit 13 (S7). The result transmitted by the analysis unit 12 may be any of the classification result of the classification unit 12a, the analysis result of the time series analysis unit 12b, and the calculation result of the index calculation unit 12c, or may be all. The storage unit 13 adds those data to the past data, combines the data for a certain period in the past, and transmits the data to the analysis unit 12 (S8). The analysis unit 12 analyzes the data transmitted from the storage unit 13 for a certain period in the past, and transmits the result to the display content determination unit 14 (S9). The content transmitted by the analysis unit 12 to the display content determination unit 14 may be any of the classification result of the classification unit 12a, the analysis result of the time series analysis unit 12b, and the calculation result of the index calculation unit 12c, or may be all. Further, when the display item is specified by the feedback from the user or the like, the analysis unit 12 may acquire the item from the display content determination unit 14 in advance and determine the transmission content.

If the display content on the terminal device 50 is preset (S10), the display content determination unit 14 converts the data received from the analysis unit 12 according to the set display content (S11), and converts the converted data. The output unit 15 transmits to the terminal device 50 (S12). The timing of output by the output unit 15 and the timing of display by the display unit 18 may be changed for each user of the behavior analysis system 1. For example, if the user is an individual, it will be output or displayed on request, if the user is an insurer, it will always be output or displayed, and if the user is a day service, it will be set regularly for visit dates and detection of abnormal behavior. Etc. may be output or displayed as a trigger.

The user may return feedback to the behavior analysis system 1 using the input unit 17 of the terminal device 50 with respect to the result displayed by the display unit 18 of the terminal device 50 (S13). When there is feedback, the display content determination unit 14 of the behavior analysis system 1 reflects it in the setting of the display content. If there is no feedback, the process ends (S14). The loop may be configured to automatically start again (S1) after the end (S14).

In the following, the operation of the system will be specifically described when the user of the behavior analysis system 1 is assumed to be an individual who wants to watch over an elderly parent. First, an individual who wants to watch over an elderly parent subscribes to a watching service using the behavior analysis system 1. After joining, the motion sensor 31, illuminance sensor 32, temperature / humidity sensor 33, radio wave sensor 34, image sensor 35, gateway and home appliances equipped with sensors, etc., which are necessary for watching over, are individuals or watching over elderly parents. It will be sent to the target parent. Then, an individual who wants to watch over the elderly parent or the parent himself installs these sensors in the home of the parent who is the analysis target.

Of the sensors installed, the motion sensor 31 and the radio wave sensor 34 have the same effect in detecting the movement of the body of the person to be analyzed, but in general, the time interval of measurement and the range of measured values are different. Different information can be obtained from the same measurement target. For example, what the motion sensor 31 detects is limited to the presence / absence of humans, while it is inexpensive, requires a small amount of power for operation, and requires a small amount of data. It can always be measured from the viewpoint. On the other hand, the radio wave sensor 34 can estimate biometric information and motion information based on minute vibrations on the body surface by setting the measurement time interval to 1 second or less, but it is expensive and has a constant power for operation. In addition, since the data capacity is relatively large, the installation location is limited and it may be difficult to operate at all times. Therefore, in this embodiment, by using these in combination, the advantages of both are utilized, and while constantly measuring from multiple viewpoints, detailed biological information and exercise information are measured and analyzed in a specific situation, time, and place. ..

Further, by simultaneously installing an illuminance sensor 32, a temperature / humidity sensor 33, etc. that measure the environment in which the analysis target is placed, in addition to the motion sensor 31 and the radio wave sensor 34, the room can be moved even with the same degree of body movement. It can be classified when it behaves differently depending on the brightness, temperature, and humidity. For example, reading a book in bed at night and sleeping can be classified as different behaviors.

Furthermore, by mounting the sensor on home appliances, the sensor can be installed in the user's home with less resistance than the sensor alone. For example, by using the data of the door open / close sensor mounted on the refrigerator, it becomes easier for the analysis subject to accept it without feeling any resistance, rather than installing another new sensor. The type of sensor used may be changed according to the behavior or service plan that you want to watch.

After installing the sensor, secure the power supply by power connection, battery, or energy harvesting such as sunlight or vibration, and start measurement. Sensor data is sent to the server either directly or via the gateway. This server may be a server of a company that provides a watching service, or may be a server that can be rented in general.

FIG. 3 is a diagram for explaining that the behavior analysis system 1 of this embodiment classifies the behavior of the analysis target person using different types of sensors and outputs the characteristics of the behavior. As shown in FIG. 3, the sensors are two types of sensors having different measurement intervals, that is, a sensor that detects a reaction discretely such as a refrigerator door open / close sensor, a motion sensor 31, and an illuminance sensor 32, and a millimeter wave sensor. It is divided into sensors that react continuously, such as.

The analysis unit 12 obtains the amount of activity of the person to be analyzed based on the measurement data of the 24-hour millimeter-wave sensor acquired by the information acquisition unit 11 and stored in the storage unit 13. Further, the analysis unit 12 estimates (classifies) what the behavior corresponding to each of the obtained activity amounts is based on the measurement data of the sensor that reacts discretely such as the motion sensor 31. The estimated result is assigned a unique identification symbol such as "sleep", "meal", and "relaxation" by the classification unit 12a. For example, when the motion sensor 31 in the living room reacts, it is presumed to be "relaxed".

The analysis unit 12 also extracts the features found in each estimated behavior. For example, when the information acquisition unit 11 acquires information on the use of the microwave oven and the opening / closing of the refrigerator from the sensor of the microwave oven or the refrigerator, or the information on the reaction of the motion sensor 31 in the kitchen, the analysis unit 12 causes the cooking activity. Extract that there was. In addition, the analysis unit 12 can extract how many times the toilet has been moved based on the amount of activity at night (during sleep) acquired from the millimeter wave sensor. Further, the analysis unit 12 extracts the agility of the action during relaxation from the movement speed during relaxation acquired from the millimeter wave sensor, and extracts the toilet movement speed at night (during sleep) acquired from the millimeter wave sensor. You can do it.

The behavioral characteristics such as the time and number of actions such as "sleep" and "meal" analyzed by the analysis unit 12 or the number of cooking activities and toilets are accumulated in the storage unit 13 for a certain period of time, and if necessary, long-term behavioral tendencies. Is transmitted to the terminal device 50 by the output unit 15. Therefore, the user can grasp the life pattern of the analysis target person. Further, the time-series analysis unit 12b may analyze how the behavioral tendency changes every day and transmit it as time-series information from the output unit 15 to the terminal device 50. Further, the time-series analysis unit 12b can predict how the behavioral tendency will change in the future, detect an abnormal sign, and transmit the result from the output unit 15 to the terminal device 50. be.

Here, the output unit 15 provides information on whether or not the estimation result to which the same identification symbol is assigned is periodically generated in a predetermined period and a specific time zone, for example, from 12:00 in the past month. Information such as whether you are having lunch at 13:00 or sleeping from 22:00 to 6:00 may be output. As a result, the user can grasp whether the analysis target person is eating regularly, whether the analysis target person's life rhythm is constant, and the like. The time zone is not limited to a predetermined time zone, and it is also possible to include an error range of about 30 minutes to 1 hour in the average of the occurrence time.

In addition, the output unit 15 is the number of times that a specific estimation result occurs within a predetermined period, for example, the number of times of going out in a day, the number of times of exercising such as gymnastics in a day, and general household chores such as cooking, washing, and cleaning. You may output the number of times of. As a result, the user can grasp whether or not the analysis target person takes various actions and lives well.

Further, the output unit 15 may output the total time during which a specific estimation result occurs within a predetermined period, for example, one day's sleep time, one day's outing time, and the like. As a result, the user can grasp whether the analysis target person has sufficient sleep time, and whether the analysis target person has sufficient walking time or time to interact with others.

Alternatively, the output unit 15 may output the number of times the data belonging to the specific estimation result changes within a predetermined period, for example, the number of times the motion sensor 31 reacts during relaxation. Even during the same relaxation, when the number of reactions of the motion sensor 31 is large, it can be seen that the movement of the analysis target person is active.

The information acquisition unit 11 of the behavior analysis system 1 acquires the sensor data collected in the server in the behavior analysis system 1. The company that provides the monitoring service may unify the format of the sensor data in the conversion unit 16 if necessary according to the type of analysis performed by the analysis unit 12. For example, if you want to grasp the rough behavior of the day in a schedule format, the data capacity is too large if you acquire the sensor data in 1-second increments, but the particle size of the information is insufficient in 1-hour increments. May be unified to. Alternatively, if you want to detect a fall in seconds, but you only need to know when you go out on a daily basis, data such as a radio wave sensor 34 that can detect a fall is in 1 second increments, and a motion sensor 31 or a door open / close sensor that can detect going out, etc. The data may be converted in 1 minute or 1 hour increments. Next, the information acquisition unit 11 or the conversion unit 16 transmits the current sensor data to the analysis unit 12, and the analysis unit 12 analyzes the current data.

For example, when it is desired to monitor the frequency of going out of the analysis target person, an algorithm that can classify going out based on the sensor data is constructed in the analysis unit 12, and the analysis unit 12 determines whether or not the current action is out of the office by this algorithm. Classify. After that, the analysis unit 12 transmits the classification result of the current sensor data to the storage unit 13 and adds it to the past sensor data. Further, the storage unit 13 extracts the sensor data for a certain period in the past including the current sensor data, and transmits the sensor data to the analysis unit 12. The analysis unit 12 can extract time-series changes in the frequency of going out by analyzing the sensor data for a certain period in the past including the current sensor data. The analysis result by the analysis unit 12 is transmitted from the output unit 15 to the user's terminal device 50 according to the setting of the display content determination unit 14. For example, the setting of the display content determination unit 14 is such that the number of times of going out in the last 3 months is averaged and output every week, the interval of going out in the last month is graphed, and the frequency of going out in the next month is predicted. However, these periods are examples and may be changed to any period.

In addition, although going out is explained as an example of the action to watch over here, it may be an action to watch over actions other than going out including general household chores such as sleeping, relaxing, eating, cooking, cleaning, washing, etc., and watching over by combining multiple of these. May be. Further, a means such as detecting an abnormal behavior and raising an alert may be used. Here, the abnormal behavior may be defined in advance by defining the type of behavior, or may be defined based on the frequency of occurrence and the probability of occurrence.

Individuals who want to watch over their elderly parents can intervene in their elderly parents based on the display results of the time-series information of these behaviors. Specifically, if an abnormality exceeding a predetermined threshold is detected, the person rushes to the facility, and if there is a sign that the behavior changes, the frequency of contact is increased because the health condition may have changed.

In addition, an individual who wants to watch over an elderly parent can return feedback from the input unit 17 of the terminal device 50 to the displayed result. For example, although multiple types of results are displayed at first, each result is given a score in 10 steps from 1 to 10 and feedback is given, so that the results with higher scores will be narrowed down and displayed from the next time onward. be.

In the following, the operation of the system when the user of the behavior analysis system 1 is assumed to be an insurer will be specifically described. The content that overlaps with Example 1 is omitted, and only the different parts are described. First, the insurer prepares an insurance service linked with the behavior analysis system 1. When there is an application for subscription to the insurance service, sensors are sent to the subscriber's home as in the first embodiment. The type of sensor used may be changed according to the insurance service plan. The sensor starts the measurement as in the first embodiment, and the sensor data is transmitted to the server. The information acquisition unit 11 of the behavior analysis system 1 acquires the sensor data collected in the server in the behavior analysis system 1, and the conversion unit 16 unifies the format of the sensor data as needed.

When an insurer utilizes this system, the setting of the display content determination unit 14 may be, for example, the total time of each human action, the difference from the previous display, the future forecast, etc., in addition to the content of the first embodiment. .. The insurer can delay the coverage of insurance by encouraging the subscriber to change his / her behavior based on these display results and improving his / her health condition. Specifically, if sleep time is insufficient, it is recommended to secure more sleep time. As a result, if the health condition of the subscriber is improved, the insurance premium can be reduced and the competitiveness of the insurance service can be improved. The insurer can give feedback on the displayed result.

In the following, the operation of the system when the user of the behavior analysis system 1 is assumed to be a day service provider will be specifically described. The content that overlaps with Example 1 is omitted, and only the different parts are described. First, the day service provider prepares a day service plan linked with the behavior analysis system 1. The day service plan includes, for example, setting up a special backup system and using the output of the behavior analysis system 1 as a reference for daily visits. When there is an application for subscription to the day service, the sensors are sent to the subscriber's home as in the first embodiment. The type of sensor used may be changed according to the day service plan. The sensor starts the measurement as in the first embodiment, and the sensor data is transmitted to the server. The information acquisition unit 11 of the behavior analysis system 1 acquires the sensor data collected in the server in the behavior analysis system 1, and the conversion unit 16 unifies the format of the sensor data as needed.

When a day service provider utilizes this system, the setting of the display content determination unit 14 is, for example, in addition to the contents of Examples 1 and 2, the schedule format data that displays daily activities in chronological order and is visualized. be. In this case, the display content determination unit 14 arranges the identification symbols (“sleep”, “meal”, etc.) assigned by the classification unit 12a in chronological order, and the output unit 15 transmits the schedule format data to the terminal device 50. The day service provider can refer to the data in the schedule format and make a plan to visit the subscriber at a time when there is a high possibility of being at home, which improves work efficiency. The day service provider can give feedback on the displayed result.
[Modification example]

The behavior analysis system 1 may be different from the first configuration described in the above embodiment. FIG. 4 is a block diagram showing a second configuration of the behavior analysis system 1. In the behavior analysis system 1 of this modification, a part or all of the classification unit 12a, the time series analysis unit 12b, the storage unit 13, the display content determination unit 14, the output unit 15, and the conversion unit 16 of the analysis unit 12 perform behavior analysis. It exists in an external server 21 provided outside the system 1 or a terminal device 50 owned by the user. In the case of this configuration, in addition to acquiring sensor data, the information acquisition unit 11 transmits data to the external server 21 in which the conversion unit 16 is installed, the classification unit 12a, the time series analysis unit 12b, and the external unit in which the storage unit 13 is installed. Data is transmitted to and received from the server 21 and the user's terminal device 50. Further, the external servers 21 may send and receive information as needed.

FIG. 5 is a diagram showing a trader having a behavior analysis system 1 and a related party. The trader 100 having the behavior analysis system 1 can send and receive information to and from the partner company 101, the partner municipality 102, the partner organization 103, and the user 104. For example, while disclosing information 110 regarding daily life to these related parties to the extent that the consent of the user 104 is obtained, information 111 regarding products and services from the partner company 101, information 112 about the regional event from the partner municipality 102, and information 112 from the partner organization. The latest academic information 113 may be acquired from 103, and the information 114 may be displayed on the terminal device 50 of the user 104 determined to be necessary to promote behavior change.

1: Behavior analysis system, 11: Information acquisition unit, 12: Analysis unit, 12a: Classification unit, 12b: Time series analysis unit, 12c: Index calculation unit, 13: Accumulation unit, 14: Display content determination unit, 15: Output Unit, 16: conversion unit, 17 input unit, 18 display unit, 21: external server, 31: human sensor, 32: illuminance sensor, 33: temperature / humidity sensor, 34: radio wave sensor, 35: image sensor, 50: terminal Equipment, 100: Contractor with behavior analysis system, 101: Partner company, 102: Partner municipality, 103: Partner organization, 104: User, 110: Information on daily life, 111: Information on products and services, 112: For local events Information, 113: Latest academic information

Claims

It is a behavior analysis system equipped with an information acquisition unit, an analysis unit, and an output unit.
The information acquisition unit acquires data from at least one sensor that captures changes caused by the behavior of the person to be analyzed.
The analysis unit classifies the data acquired by the information acquisition unit into one or more types, and assigns a unique identification symbol to at least one of the classification results.
The output unit is a behavior analysis system characterized by outputting time-series information of classification results to which the same identification symbol is assigned.
In the behavior analysis system according to claim 1,
The information acquisition unit is characterized by acquiring data of at least one sensor that detects the movement of the analysis target person and data of at least one sensor that measures the environment in which the analysis target person is placed. Behavior analysis system.
In the behavior analysis system according to claim 1,
The information acquisition unit is a behavior analysis system characterized by acquiring data from at least two sensors having different measurement intervals.
In the behavior analysis system according to any one of claims 1 to 3.
A behavioral analysis system characterized in that at least one of the sensors is mounted on a home appliance.
In the behavior analysis system according to any one of claims 1 to 3.
The output unit is a behavior analysis system characterized in that it outputs information as to whether or not a classification result to which the same identification symbol is assigned is periodically generated in a specific time zone for a predetermined period.
In the behavior analysis system according to any one of claims 1 to 3.
The output unit is a behavior analysis system characterized by outputting the number of times a classification result to which the same identification symbol is assigned occurs within a predetermined period.
In the behavior analysis system according to any one of claims 1 to 3.
The output unit is a behavior analysis system characterized in that the classification result to which the same identification symbol is assigned outputs the total time generated within a predetermined period.
In the behavior analysis system according to any one of claims 1 to 3.
The output unit is a behavior analysis system characterized in that data belonging to a classification result to which the same identification symbol is assigned outputs the number of times the data has changed within a predetermined period.
In the behavior analysis system according to any one of claims 1 to 3.
The output unit is a behavior analysis system characterized in that the identification symbols are output in a schedule format arranged in chronological order.
A step in which the information acquisition unit acquires data from at least one sensor that captures changes caused by the behavior of the person to be analyzed.
A step in which the analysis unit classifies the data acquired by the information acquisition unit into one or more types and assigns a unique identification symbol to at least one of the classification results.
The output unit outputs the time-series information of the classification result to which the same identification symbol is assigned, and
Behavioral analytics method.