WO2023112757A1

WO2023112757A1 - Sensing system, sensing device, and sensing method

Info

Publication number: WO2023112757A1
Application number: PCT/JP2022/044782
Authority: WO
Inventors: 健太郎佐野; 佐知田中; 昭義大平; シャヘッドサルワル; 景子藤咲
Original assignee: 株式会社日立製作所
Priority date: 2021-12-17
Filing date: 2022-12-05
Publication date: 2023-06-22
Also published as: JP2023090273A

Abstract

This sensing system (100) comprises a plurality of sensors (10), including at least one human-motion sensor. A space in which the sensors are installed is divided into a plurality of areas. A plurality of the sensors are installed in the plurality of areas respectively. As a subject moves, sensor information obtained by converting discrete data of at least one of the sensors into continuous data is acquired as a degree of importance. The behavior of the subject is recognized on the basis of a weight calculated using information from the plurality of sensors including the degree of importance, and the information from the sensors is utilized to suitably recognize human behavior.

Description

SENSING SYSTEM, SENSING DEVICE, AND SENSING METHOD

The present invention relates to a sensing system, a sensing device, and a sensing method.

In recent years, communication technology and sensor technology have developed, and in everyday life, the results of sensing are used to recognize human movements and actions, and consideration is being given to using them for monitoring.

For example, Patent Literature 1 describes that "a watching device in a target person's house detects movement when a human sensor (human body detection means) other than the one that has responded immediately before senses a movement. When there is a response from another human sensor before the measured value t reaches the movement fixed time T, the measured value t is reset and the measurement is restarted.Response of one human sensor After the measurement is started by, when the measured value t reaches the movement fixed time T, it is determined that the person being watched has moved to the watching area of the one human sensor. According to this technology, it is possible to reduce the amount of information to be stored by specifying an important motion sensor without interfering with grasping the daily activities of the person being watched over.

JP-A-2003-132462

According to the prior art described above, the movement of a person is confirmed using the movement confirmation time T when a plurality of human sensors react. However, since lifestyle habits differ from person to person and may change according to life stages, it is difficult to appropriately set the fixed movement time T in advance.

The purpose of the present invention is to build a system that appropriately recognizes human behavior by utilizing sensor information.

In order to solve the above problems, the sensing system of the present invention is a sensing system comprising a plurality of sensors including at least one human sensor, wherein the space in which the sensors are installed is divided into a plurality of areas. A plurality of sensors are installed in each of a plurality of areas, and as the subject moves, information of the sensors is obtained by converting discrete data of at least one of the sensors into continuous data as the degree of importance. and the action recognition of the person to be measured is performed based on the weight calculated using the information of the plurality of sensors including the degree of importance.

According to the present invention, it is possible to build a system that appropriately recognizes human behavior by utilizing sensor information.

1 is a block diagram showing the configuration of a sensing system according to an embodiment; FIG. 3 is a block diagram showing the internal configuration of a computing unit; FIG. FIG. 4 is a diagram showing a case where sensor data are discrete values at time t; FIG. 10 is a diagram showing creation of curve data m1 to m5 for discrete values of reaction times t1 to t5; FIG. 3C is a diagram in which overlapping sections of curve data m1 to m5 in FIG. 3B are integrated. FIG. 9 is a diagram showing an example of applying linear approximation to discrete values of reaction times t1 to t5; FIG. 10 is a diagram showing an example of converting discrete values of reaction times t1 to t5 into rectangular functions; FIG. 4 is a diagram showing an example of randomizing the binary sensor data 18 at the time t when the sensor 10 reacts. It is a figure which shows adding the function value of less than the maximum value of sensor data in the spatial direction centering on the sensor which reacted, and making it sensor information. FIG. 4 is a plan view showing a specific example of a space to which sensor information is added in a spatial direction; It is a figure which shows the time change of the importance of a 1st sensor. Time change of the sensor information of the second sensor FIG. 10 is a diagram showing changes over time in weights calculated by an importance determining unit 102a; FIG. 10 is a diagram showing the time evolution of the importance indicated by the curve of the third sensor; FIG. 10 is a diagram showing the time evolution of importance indicated using a rectangular function of the fourth sensor; FIG. 8C is a diagram showing temporal changes in weights calculated by another importance determining unit 102b. FIG. 4 is a flow chart explaining the operation of the sensing system; FIG. 11 is a block diagram showing the configuration of a sensing system according to another embodiment; FIG. It is the figure which showed the example of arrangement|positioning in the house of the sensor 10 of this embodiment. It is a figure which shows an example of the display content of the display part in a watching service.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. However, the present invention is not limited to the following embodiments, and various modifications and applications can be made within the technical concept of the present invention. Included in the scope.

FIG. 1 is a block diagram showing the configuration of a sensing system 100 according to an embodiment.
The sensing system 100 includes an information acquisition unit 101 , a calculation unit 102 , a storage unit 103 , a display unit 104 and a sensor 10 . In this specification, the configuration of the sensing system 100 excluding the sensor 10 (the information acquisition unit 101, the calculation unit 102, the storage unit 103, and the display unit 104) is called a sensing device.

More specifically, a computer comprising a CPU that performs arithmetic processing, a memory, a camera, a communication unit, an operation unit, a display unit, and a nonvolatile storage medium executes a program stored in the nonvolatile storage medium. , the information acquisition unit 101 , the calculation unit 102 , and the display unit 104 , and constitute the sensing system 100 . The storage unit 103 is configured with a memory or a nonvolatile storage medium.

The information acquisition unit 101 acquires the sensor data 18 from the sensor 10 and also acquires the body information 105 of the person to be analyzed.
Here, the physical information 105 of the person to be analyzed includes, for example, height, weight, date of birth, BMI (body mass index), body fat percentage, visceral fat level, muscle mass, body water percentage, and body age.

The information acquisition unit 101 acquires sensor data 18 by directly connecting to the sensor 10 via Ethernet, wireless communication, or the like. Also, when collecting data from a locally installed PC connected to the sensor 10 via a gateway, the information acquisition unit 101 accesses the PC via a local network or the Internet to acquire the sensor data 18. . Furthermore, when the sensor data 18 is collected directly or via a gateway to a cloud computing service server, the information acquisition unit 101 may acquire the sensor data 18 by accessing the server via the Internet.

The information acquisition unit 101 receives sensor data 18 from an external server in a file format such as csv (Comma Separated Value) using Python (registered trademark), java (registered trademark), c language, or the like.

In addition, the information acquiring unit 101 refers to the storage unit 103 and acquires the physical information 105 of the analysis target, which is stored in the storage unit 103 by the system administrator, based on the application documents that the analysis target fills out when applying for the service. do.

The computing unit 102 analyzes a plurality of sensor data 18 and recognizes human behavior. Specifically, the calculation unit 102 classifies one or more types of sensor data 18 into one or more types. Here, the classification means grouping the sensor data 18 based on their characteristics. Assign label names that indicate human actions, such as "", "cleaning", "washing", etc.
The configuration of the calculation unit 102 will be described later.

The accumulation unit 103 accumulates at least one of the information acquired by the information acquisition unit 101 and the output of the calculation unit 102 . In addition to accumulating data for each person to be analyzed, the data may be tagged with attributes such as age, gender, origin, language, religion, and tastes and preferences of the person to be analyzed.

Further, the storage unit 103 may transmit the history data stored for the person to be analyzed to the calculation unit 102, and the calculation unit 102 may perform time-series analysis of these data. Here, in this specification, time series analysis refers to visualization of various data by graphs with time on the horizontal axis, and data values, the number of classifications, or the duration of each classification, in the time direction It is defined as calculating the rate of change, moving average, variance, standard deviation, etc., and performing operations such as error analysis and polynomial approximation. These time-series analyzes may include comparison with average values of persons belonging to the same category in terms of age, gender, origin, religion, lifestyle, occupation, medical history, and the like.

The display unit 104 receives information from the calculation unit 102 and the storage unit 103, and displays the information to the administrator of the sensing system 100, system users, analysis subjects, and the like. The display unit 104 displays numbers, characters, tables, graphs, and the like in arbitrary formats.

Next, the operation of the calculation unit 102 will be described in detail.
FIG. 2 is a block diagram showing the internal configuration of the arithmetic unit 102. As shown in FIG.
The calculation unit 102 is composed of an importance determination unit 102a, a weight determination unit 102b, and a classification unit 102c.

The importance determination unit 102a calculates the importance of the sensor data 18 acquired by the calculation unit 102. Here, in this specification, the degree of importance is calculated using the response (change) of the sensor data 18 in the temporal direction and the spatial direction. It is defined as an index that gives a clue as to whether the data 18 is relatively emphasized. A method for calculating the degree of importance will be described in detail below.

First, a case will be described where the importance determination unit 102a inputs a discrete value (for example, 1) at the time t when the sensor 10 reacts to the acquired sensor data 18, as shown in FIG. 3A, as sensor information.

Note that in FIG. 3A, the sensor 10 is an example of a human sensor (pyroelectric sensor) using a pyroelectric effect. A human sensor that uses the pyroelectric effect will detect when a person stops moving for a certain period of time even within the detection range, or when a person leaves the detection range, that is, when there is no temperature change for a certain period of time. value will be 0. FIG. 3A shows temporal changes in sensor information from one human sensor. In FIG. 3A, the human sensor does not react between reaction time t1 and reaction time t2. It will be in a state of not doing. This point is the same between reaction time t2 and reaction time t3 and between reaction time t4 and reaction time t5. Between reaction time t3 and reaction time t4, the person to be analyzed stops moving, or leaves the detection range of the human sensor after reaction time t3, and returns to the detection range of the human sensor at reaction time t4. It is assumed that the

When the discrete value sensor information of FIG. 3A is input, the importance determination unit 102a sets the function value at time t to the maximum value (for example, 1) as shown in FIG. Create sensor information with added function value less than value. Specifically, the importance determining unit 102a creates curve data m1 to m5 for discrete values (eg, 1) of the sensor data 18 at reaction times t1 to t5.

Here, the function value less than the maximum value is calculated by the importance determination unit 102a so that the function value decreases as the time difference from time t increases. For example, the importance determination unit 102a determines a function value that is less than the maximum value in the time direction before and after the sensor reaction time t, a normal distribution, a Student's τ distribution, a U (Universal) distribution, and any other arbitrary distribution used in the statistical field. apply the distribution of

When the sensor 10 is a human sensor, the importance determining unit 102a sets the human existence probability to 1.0 when the sensor information (the value of the sensor data 18) is 1, and determines the continuous values of the created sensor information to be 1.0. , can be treated as the existence probability of humans in the time series direction.

FIG. 3C is a diagram that integrates overlapping sections of the curve data m1 to m5 in FIG. 3B. If the created curve data overlap, the importance determining unit 102a integrates the overlapped curves into a curve having the maximum value. Alternatively, the importance determination unit 102a adopts the sum of a plurality of curves and normalizes it by the maximum value of the entire sensor information. Thereby, the importance determining unit 102a can uniquely determine the value of the sensor information at each time.

The sensor information in FIG. 3C is characterized by having a large value (area) so that the sensor 10 responds continuously, except when the elapsed time from the reaction time of the sensor 10 is short. In addition, it is possible to use it as an indicator of which sensor is to be emphasized.
In this specification, as shown in FIG. 3C, the sensor information converted so as to be continuously distributed in the time direction is defined as importance.

The importance determining unit 102a converts discrete sensor data 18 into continuous sensor data 18 continuously distributed in the time direction, as shown in FIGS. 3B and 3C. As a result, it is possible to prevent the value from suddenly becoming 0 or invalid due to an unintended loss of the sensor data 18, so that a system that is robust against loss of the sensor data 18 can be realized.

In the above, it was explained that the importance determination unit 102a converts the discrete sensor data 18 into continuous values having an arbitrary distribution used in the statistical field, as shown in FIG. 3B. However, other conversion methods may be used.

FIG. 4A shows that when the importance determination unit 102a adds function values less than the maximum value in the time direction before and after the sensor reaction time t to convert discrete values into continuous values, the discrete values of the reaction times t1 to t5 It is a figure which shows the example which changed to the curve and applied linear approximation with respect to a value. Compared to the case of FIG. 3B, the amount of calculation of the importance determining unit 102a can be reduced.

FIG. 4B shows an example in which the importance determining unit 102a converts the discrete values of the reaction times t1 to t5, which are the sensor data 18 at the time t when the sensor 10 reacts, into a rectangular function. This is effective for sensor data 18 whose value is approximately constant for a short period of time, such as from a sensor.

FIG. 4C is an example in which the importance determination unit 102a randomizes the binary sensor data 18 at the time t when the sensor 10 reacts (for example, 1 or 0 for a sensor with a maximum value of 1 and a minimum value of 0). is.

As a more specific setting method for randomization, the possible range of random values is changed for the minimum value (eg 0) and the maximum value (eg 1) of the discrete values of the sensor data 18 respectively. For example, if the sensor data 18 is "0", it is converted to "a random number in the range of 0 to 0.3", and if it is "1", it is converted to "a random number in the range of 0.7 to 1.0 Random value of ”.

As a result, it is possible to avoid abnormalities during analysis caused by changes in the values of the sensor data 18, and to stably drive the system.
Specifically, it prevents division of real numbers by 0 during calculation, improves the generalization performance of machine learning by distributing discrete values in a certain range, or prevents sudden changes from the minimum value to the maximum value in calculations. can be expected to have the effect of stabilizing the

Furthermore, as shown in FIG. 5, the importance determining unit 102a not only adds values in the direction of the time axis to obtain sensor information, but also adds the maximum value of the sensor data 18 in the spatial direction centering on the sensor 10 that has responded. Add a function value of less than the sensor information.

Specifically, when the sensor responds in the living room, the importance determining unit 102a may add sensor information values to the bedroom and kitchen that are spatially adjacent to the living room in a movable manner.

FIG. 6 is a plan view showing a specific example of a space to which the sensor information shown in FIG. 5 is applied. The importance determining unit 102a adds a relatively large value (existence probability: 0.7) to the nearest space, the kitchen and the corridor, starting from the living room (existence probability: 1.0) where the sensor data 18 is generated, and moves away from the living room. Decrease the value to be added.

Next, the weight determination unit 102b in FIG. 2 will be described. The weight determination unit 102b converts the importance determined by the importance determination unit 102a into a weight. Here, in this specification, the weight is defined as an index indicating which sensor 10 data is relatively emphasized when analyzing multiple types of sensor data 18 . For the weight, the degree of importance may be used as it is, but the degree of importance is converted by a plurality of methods described in detail below.

7A, 7B, and 7C, a first example in which the weight determination unit 102b converts the importance into a weight based on a combination of a plurality of sensors. FIG. It is a figure which shows the time change of the importance of a sensor.

The weight determination unit 102b uses the time change of the sensor information (sensor data 18) of the second sensor shown in FIG. Calculate the weight by integrating. FIG. 7C is a diagram showing changes over time in weights calculated by the importance determining unit 102a.

More specifically, the first sensor is, for example, a human sensor installed in the kitchen. The importance determining unit 102a calculates the importance with the maximum value being "1", which can be regarded as the existence probability of a person.

The second sensor is, for example, a door open/close sensor. The second sensor, which is a door open/close sensor, is pre-installed in the connected home appliance, or is externally attached to the refrigerator or microwave oven. The second sensor measures sensor information indicating the open/close status of the doors of refrigerators, microwave ovens, and the like.

Continuing with the description of the weights, the graph of the time change of the weights calculated by the weight determination unit 102b in FIG. Indicates that there is a high possibility that

Therefore, by recognizing human behavior based on the time change of weights as shown in FIG. 7C, human behavior can be recognized with higher accuracy than when based only on the data of the human sensor. In particular, it functions effectively when actions such as going to the kitchen to wash hands during a meal or washing dishes in the kitchen after a meal are distinguished from cooking.

It has been described above that the weight determination unit 102b calculates the weight by integrating the importance of the first sensor and the sensor information of the second sensor. If there is, the weight may be used as the weight of the sensor information of the second sensor, and if the importance of the first sensor is less than 0.5, the weight may be ignored (the weight value may be set to 0).

Next, a second example in which the weight determination unit 102b converts the importance into a weight by combining a plurality of sensors will be described with reference to FIGS. 8A, 8B, and 8C.

FIG. 8A is a diagram showing a temporal change in the degree of importance indicated by the curve of the third sensor determined by the degree-of-importance determining unit 102a.
FIG. 8B is a diagram showing changes over time in importance indicated by using the rectangular function of the fourth sensor determined by the importance determining unit 102a.

The weight determining unit 102b calculates the weight by integrating the temporal change in importance of the third sensor in FIG. 8A and the temporal change in importance of the fourth sensor in FIG. 8B. FIG. 8C is a diagram showing changes over time in weights calculated by the importance determining unit 102a.

More specifically, the third sensor is, for example, a human sensor installed in the bedroom. The importance determining unit 102a calculates the importance with the maximum value being "1", which can be regarded as the existence probability of a person.

The fourth sensor is, for example, an illuminance sensor in the bedroom, which measures the illuminance in the room as sensor information by being pre-installed in the lighting or installed in the room as an afterthought. Since the measured value of the illuminance sensor is considered to be a constant value for a short period of time, it is appropriate to calculate the degree of importance using a rectangular function.

Furthermore, in the case of the illuminance sensor, the degree of importance is set to a certain threshold value (for example, 10 [lx]) in order to express the binary state of ON/OFF of the lighting, and if it is equal to or higher than the threshold value, ON is set. It may be "1" to indicate the value, and "0" to indicate OFF when the value is equal to or less than the threshold.

In this case, the graph of FIG. 8C showing changes in weight over time calculated by the weight determining unit 102b indicates that the greater the weight, the higher the possibility that a person is present in the bedroom and active.

Therefore, by recognizing human actions based on the weights shown in FIG. 8C, human actions can be recognized with higher accuracy than when based only on the data of the human sensor. In particular, it works effectively when you are in your bedroom but not asleep and are doing activities such as relaxing or reading, and distinguishing them from sleep.

It has been described above that the weight determining unit 102b calculates the weight by integrating the temporal change in the importance of the third sensor and the temporal change in the importance of the fourth sensor. If each of the degrees of importance of the fourth sensor is equal to or greater than a predetermined value, the weight is calculated by integration, and if not, the weight is ignored (the weight value is set to 0). good.

Here, returning to FIG. 2, the classification unit 102c will be described.
The classification unit 102c classifies human behavior based on the weights determined by the weight determination unit 102b. Algorithms used for classification are not particularly limited.

For example, the classification unit 102c may set a threshold for each weight, and identify the corresponding action when the weight is greater than or equal to the threshold. Specifically, in FIG. 8C, if the weight has a value of 0.8 or greater, the person is classified as relaxing in the bedroom.

In addition, the classification unit 102c uses supervised machine learning to prepare teacher data and learning data including cases to be classified in advance, train a classifier capable of classifying data according to the number of types of teacher data, and train Human behavior may be classified by a classifier based on
Further, the classification unit 102c may learn the physical information 105 of the person to be analyzed and prepare a classifier for each person to be analyzed.

Here, an example of teacher data is human behavior such as "sleep" and "going out", and an example of learning data is sensor data 18 of a time period corresponding to "sleep" and "going out". The sensor data 18 may be converted into importance or weight and used as learning data.

The type of machine learning is not limited as long as it learns using teacher data and acquires a classification function. For example, methods using decision trees including boosting, logistic regression, k nearest neighbors, support vector machines, random forests and their ensembles, fully connected layers of deep learning, CNN, classifiers constructed using RNN, etc. can be utilized.

Further, the classification unit 102c may collect data having similar feature amounts of the collected sensor data 18 by using, for example, unsupervised learning as a classification method.

The type of machine learning is not limited as long as it has a classification function without teacher data, such as k-means method, k-means++ method, x-means method, k-shape method, mixed Gaussian model, etc. clustering methods, methods using anomaly detection methods such as OneClass SVM, Elliptic Envelope, Isolation Forest, Local Outlier Factor, mixed Gaussian models, and clustering methods that combine them with dimensionality reduction methods such as t-sne and autoencoder , Adversarial Autoencoders, and the like, which uses the low-dimensional space of adversarial generative networks and collects data into a predetermined distribution.

In addition, the method of determining the number of classifications arbitrarily, the method of determining an appropriate value using methods such as the elbow method and the silhouette method, and the method of defining a distance index between feature values without predetermining it, and Any method may be used, such as counting as different classifications if there are items that differ by more than the value.

FIG. 9 is a flow diagram illustrating the operation of sensing system 100. As shown in FIG.
The sensing system 100 starts processing in response to an operation by the system administrator.

In step S1, the information acquisition unit 101 acquires the sensor data 18 of the sensor 10 (human sensor, door sensor, etc.) and the body information 105 of the person to be analyzed.

In step S2, the information acquisition unit 101 transmits the data acquired in step S1 to the calculation unit 102 and the storage unit 103.

In step S3, the calculation unit 102 uses the sensor data 18 received from the information acquisition unit 101 to perform action recognition.

Specifically, in the calculation unit 102, the importance determination unit 102a converts discrete data from the sensor data 18 of the plurality of sensors 10 into continuous data to obtain the importance, and the weight determination unit 102b calculates the importance of the plurality of A change over time of a weight that is the product of the importance of the sensor 10 and the raw sensor data 18 or the product of the importances is calculated, and the classification unit 102c recognizes actions based on the weight.

In step S4, the calculation unit 102 transmits the recognition result to the display unit 104 and the storage unit 103.

In step S5, the accumulation unit 103 transmits the recognition result notified from the calculation unit 102 in step S4 to the calculation unit 102 together with the past accumulated data of the person to be analyzed or the other person.

In step S6, the calculation unit 102 utilizes the current action recognition result obtained in step S3 and past accumulated data to calculate a time-series change and transmit it to the display unit 104.

In step S7, the display unit 104 displays the current action recognition result and its chronological change transmitted in step 6.

In step S8, the sensing system 100 determines whether or not to end the process, and may end as it is (Yes in S8), or move to step S1 and repeat the process. A loop process that returns immediately after the process start (S1) may be implemented by presetting (No in S8).

Next, with reference to FIG. 10, a case will be described in which a business operator, such as an insurance company or a day service provider, that provides services and products to the elderly utilizes the sensing system 100 of the embodiment.

The human sensor 11, the illuminance sensor 12, the temperature/humidity sensor 13, the door opening/closing sensor 14, and the acceleration sensor 15 in FIG. The sensor data 18 of the door opening/closing sensor 14 and the acceleration sensor 15 (hereinafter collectively referred to as the sensor 10) are transmitted to the external server 31 directly or via the router 16, and then the information of the sensing system 100 is acquired. It is sent to the unit 101 .

Also, the sensor 10 is installed when there is an application for subscription to a service or product provided by a business operator, and a power source is secured by power connection, battery, or energy harvesting such as sunlight or vibration, and measurement is started.

FIG. 11 is a diagram showing an example of in-home arrangement of the sensors 10 of the present embodiment.
At least two types of the human sensor 11, the illuminance sensor 12 and the temperature/humidity sensor 13 may be treated as a set as an environment sensor. In addition, a microphone (not shown), a noise sensor, an air pressure sensor, an odor sensor, a pressure sensor, a wearable sensor, a weight scale, a body composition monitor, an indoor/outdoor image sensor, an image sensor built into a robot cleaner, etc. can be incorporated into the system. good.

The home is divided into multiple areas, and a human sensor 11, illuminance sensor 12, and temperature/humidity sensor 13 are installed in each area. The area division method may be for each room. When a plurality of actions are performed in one room, the ranges in which each action is performed may be separate areas.

A door open/close sensor 14 is installed in each door, microwave oven 22, and refrigerator 23 to sense the open/close status of the door and home appliances. Acceleration sensors 15 are installed in the washing machine 21, the kitchen sink, and the robot cleaner 24 to sense the usage conditions of each.

The data of these sensors 10 are collected by the router 16, and the router 16 transmits the data to the information acquisition unit 101. In addition, the data of each sensor 10 installed in the home appliance may be replaced with the data of the connected home appliance by replacing the sensor 10 .

Here, since the human sensor 11 has a viewing range of ±45 degrees in the horizontal direction and about 5 m in the depth direction, it should be installed so as not to detect information in areas other than the area in which the human sensor 11 is installed. is desirable. For example, it is important to arrange each human sensor 11 so that the center axis 17 of the visual field range of the human sensor 11 does not substantially face the doorway to the adjacent area.

The type of sensor used can be changed according to the service and product. For example, some or all of the human sensor 11, the illuminance sensor 12, and the temperature/humidity sensor 13 are combined into one environment sensor. may It may also include data linkage with other commercially available devices such as online-connectable wearable sensors, weight scales, and body composition meters.

The external server 31 is not limited to a server owner, and may be a server managed by a business operator, a server managed by a sensor manufacturer, or a web service. It may be a server that can be rented to the general public. Furthermore, data may be stored in different environments for each business, read by the sensing system 100, and connected within the system. However, these servers need to be able to communicate with the sensing system 100 built in either the company's internal environment or the cloud environment.

The information acquisition unit 101 of the sensing system 100 acquires the sensor data 18 of the sensor 10 collected by the external server 31 and the body information 105 of the person to be analyzed into the sensing system 100 .

The body information 105 of the person to be analyzed is acquired by making it possible to enter it into the sensor 10, by making it an entry form for subscription to services and products, or making it possible to additionally enter it from my page on the company's homepage. . Separately, measurement instruments such as a grip strength meter and a body composition meter may be prepared, or may be obtained in cooperation with a third party such as another business operator, local government, non-profit organization, or the like.

The information acquisition unit 101 transmits the data of the sensor 10 and the body information 105 of the person to be analyzed to the calculation unit 102 and the storage unit 103, as described in FIG.

As described in FIG. 1, the calculation unit 102 recognizes human behavior from the information of the sensor 10 received from the information acquisition unit 101 and the body information 105 of the person to be analyzed. Here, human actions include, for example, "sleep", "going out", "relax", "cooking", "eating", "cleaning", "washing", and "others".

As described with reference to FIG. 1, the accumulation unit 103 accumulates the data of the sensor 10 and the body information 105 of the person to be analyzed, receives the body information 105 of the person to be analyzed from the information acquisition unit 101, and obtains the past information of the same person. history data such as physical information and amount of activity, and transmits the data to the information acquisition unit 101 .

The storage unit 103 may be provided in the external server 31. In this case, the information acquisition unit 101 transmits the past history data of the analysis subject in the external server 31 together with the current data to the calculation unit 102. do. Accordingly, the calculation unit 102 can perform time-series analysis.

The calculation unit 102 transmits the current action recognition result and its time-series analysis result to the accumulation unit 103 and the display unit 104 . The accumulation unit 103 organizes and accumulates the results output by the calculation unit for each person to be analyzed and for each category such as the age, sex, origin, religion, lifestyle, occupation, and medical history of the person to be analyzed.

The display unit 104 displays the received results to the system operator or business operator. The display method may differ between the system operator and the provider.
For the system operator, for example, in addition to the current behavior of the person to be analyzed and its time-series analysis results, information on the operating state of the sensor, information on the battery status of the sensor, It displays the indicated value, the date and time when the operator referred to the system, the number of times, the time and information on the content, etc.

For businesses, for example, in addition to the current behavior recognition results of the person to be analyzed and their chronological changes, it is converted into information that is particularly necessary for each business and displayed. Specifically, an insurer can infer the applicability of insurance for the current subject of analysis by simultaneously showing the history of insurance coverage of others in the past, the types of daily behavior, and their chronological changes. You may provide materials to Armed with this information, insurers can consider interventions to modify the types of daily behavior of the analyses, and their changes over time, in a way that makes insurance coverage less likely. This intervention includes, for example, making recommendations such as "Let's go out for about 30 minutes every day" and "Let's go to bed 15 minutes earlier today."

In the case of a day service provider, by simultaneously showing the schedule of the day service and changes in the types of daily activities, it would be possible to provide materials for examining the impact of the content of the day service on the lifestyle of the subject of analysis. may Based on this information, the day service provider can consider a program for correcting the chronological change in the type of daily behavior for each person to be analyzed in a desirable direction.

Businesses can give feedback to the system administrator about the displayed content. Means of feedback include verbal communication, email, and posting via the system administrator's home page. The system administrator may change the display contents based on the feedback from the operator.

The contents of the sensing system 100 of the embodiment can also be applied to individuals whose elderly parents want to watch over them. For example, an individual who wants to watch over an elderly parent may sign a contract for a watching service provided by a business, and sensors may be sent. By installing them in the elderly parent's home and sending the data to the business operator, it is possible to receive the type of daily behavior of the elderly parent as an analysis result and monitor the lifestyle habits. In addition, based on the results, it is possible to take actions such as encouraging parents who rarely go out to go out or visiting parents who wake up late.

FIG. 12 is a diagram showing an example of display contents of the display unit 104 in the monitoring service.
The display unit 104 displays the current action recognition result "cooking" recognized by the calculation unit 102, the amount of activity recognized by the action recognition, and the time-series analysis result "sleep" on the display of the terminal of the individual who wants to watch over his elderly parent. A pie chart showing the time ratio of daily activities such as "eating" and "relaxing" is displayed. Furthermore, the display unit 104 displays the chronological change in the amount of activity for each month, and displays analysis comments on the living situation.

In addition, the present invention is not limited to the above-described examples, and includes various modifications. The above embodiments have been described in detail to facilitate understanding of the present invention, and are not necessarily limited to those having all the described configurations. Also, part of the configuration of one embodiment can be replaced with the configuration of another embodiment, and the configuration of another embodiment can be added to the configuration of one embodiment.

10 sensor 18 sensor data 100 sensing system (sensing device)
101 Information acquisition unit 102 Calculation unit 102a Importance determination unit 102b Weight determination unit 102c Classification unit 103 Accumulation unit 104 Display unit 105 Physical information of person to be analyzed

Claims

A sensing system comprising a plurality of sensors including at least one human sensor,
The space in which the sensors are installed is divided into a plurality of areas, and a plurality of sensors are installed in each of the plurality of areas,
Obtaining sensor information obtained by converting discrete data of at least one sensor into continuous data as the subject moves, as an importance level;
A sensing system for recognizing the behavior of the person to be measured based on a weight calculated using information from the plurality of sensors including the degree of importance.
The sensing system according to claim 1,
Each sensor is arranged so as not to detect information in an area other than the area in which they are installed, and the central axis of the measurement range of at least one sensor does not substantially face the entrance to the adjacent area. sensing system.
The sensing system according to claim 1,
A sensing system according to claim 1, wherein the weight is obtained from the importance of the sensor and the data of the sensor.
The sensing system according to claim 1,
The sensing system, wherein the weight is obtained from the importance of two sensors.
In the sensing system according to any one of claims 1 to 4,
The degree of importance is obtained by converting temporally or spatially discrete data of the sensor into temporally or spatially continuous data using any one of curves, straight lines, rectangles, and random numbers. A sensing system characterized by:
In the sensing system according to any one of claims 1 to 4,
A sensing system characterized by performing time-series analysis of behavior recognition of the person to be measured.
A sensing method for a sensing system comprising a plurality of sensors including at least one human sensor,
Acquiring sensor data from sensors installed in each area of the space divided into a plurality of areas;
a step of obtaining sensor information obtained by converting discrete data of the sensor data into continuous data as a degree of importance;
a step of recognizing an action based on a weight calculated using information of a plurality of sensors including the importance;
A sensing method, comprising:
A sensing device that recognizes the behavior of a subject based on sensor data from a plurality of sensors including at least one human sensor,
an information acquisition unit that acquires sensor data as sensor information from sensors installed in each of the areas divided into a plurality of areas;
an importance determining unit that obtains, as an importance, sensor information obtained by converting discrete data of the sensor data into continuous data;
a weight determination unit that calculates a weight using information of a plurality of sensors including the importance;
a classification unit that recognizes the behavior of the person to be measured based on the weight;
A sensing device comprising:
The sensing device according to claim 8,
The sensing device according to claim 1, wherein the weight determining unit calculates the weight from the importance of the sensor and data of other sensors, or calculates the weight from the importance of a plurality of sensors.