JP2007052474A - Behavior identification system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a behavior identification system, capable of identifying a behavior pattern of a user through a simple structure. <P>SOLUTION: A transmitter 14 transmitting an infrared signal having a fixed pattern is installed to the rear head part of a user H who arbitrarily moves. On the other hand, light receiving modules 12a and 12b forming a receiver 12 detect the infrared signal transmitted from the transmitter 14, by observing two detection areas different in at least either of size and position. A main body 12c forming the receiver 12 forms a receiving frequency function showing the receiving frequency for unit time of the infrared signal detected by each of the modules 12a and 12b. The behavior pattern of the user H is identified by applying the features of two formed receiving frequency functions to a decision tree. According to this invention, the behavior pattern of the user can be identified through a simple structure. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a behavior identification system, and more particularly to a behavior identification system that identifies, for example, a user's behavior mode.

An example of a conventional system of this type is disclosed in Patent Document 1. According to this prior art, the lighting state of the infrared tag attached to the user is controlled by the location server. Infrared light emitted from the infrared tag is captured by a plurality of infrared cameras. The position of the infrared tag is detected by the video server based on a plurality of captured infrared images. The location server identifies the position of the user by integrating the positions of the detected infrared tags. Thereby, the position of the user can be specified with high accuracy and stability.
JP 2003-329762 A [G01S 5/16, H04N 5/225]

However, in the conventional technology, it is necessary to prepare a plurality of infrared cameras, and there is a problem that the configuration becomes complicated. In addition, the conventional technology only specifies the position of the user and cannot identify the user's behavior.
Therefore, a main object of the present invention is to provide a behavior identification system that can identify a behavior mode of a user with a simple configuration.

  Another object of the present invention is to provide an action identification system capable of detecting a user's passing direction.

  The behavior identification system (10) according to the first aspect of the present invention is a transmission means (14) that is held by a user who arbitrarily acts and sends a fixed pattern signal to each of a plurality of detection areas having at least one of size and position different from each other. A plurality of detection means (12a, 12b) for detecting the fixed pattern signal sent from the transmission means with attention, a reception frequency function indicating the reception frequency per unit time of the fixed pattern signal detected by each of the plurality of detection means Creating means (S9) for creating the user and identification means (S11, S13, S15) for identifying the user's behavior based on the plurality of reception frequency functions created by the creating means.

  The transmission means for transmitting the fixed pattern signal is held by a user who acts arbitrarily. On the other hand, the plurality of detection means detect fixed pattern signals sent from the transmission means while paying attention to a plurality of detection areas having different sizes and positions. The creation means creates a reception frequency function indicating the reception frequency per unit time of the fixed pattern signal detected by each of the plurality of detection means. The behavior mode of the user is identified by the identifying unit based on a plurality of reception frequency functions created by the creating unit.

  The plurality of detection areas noted by the plurality of detection units are different from each other with respect to at least one of size and position. For this reason, the plurality of reception frequency functions created by the creating unit depend on the user's behavior and are different from each other. A user's action mode is identified based on such a reception frequency function. That is, action identification with a simple configuration becomes possible.

  The behavior identification system according to the invention of claim 2 is dependent on claim 1, wherein the transmission means is attached to the body of the user, and the plurality of detection means are attached to different positions of the fixed object. By attaching to the body, for example, the back of the head, the detection accuracy of the fixed pattern signal differs between the front and the rear of the user. As a result, a clear difference appears in the reception frequency function, and the accuracy of action identification increases.

  The behavior identification system according to the invention of claim 3 is dependent on claim 2, wherein the fixed object is a gate, the plurality of detection means are first detection means (12a) provided on a side surface on the entrance side of the gate, and the gate. The 2nd detection means (12b) provided in the side by the side of the exit is included.

  The behavior identification system according to the invention of claim 4 is dependent on any one of claims 1 to 3, wherein the identification means is detected by a detection means (S13) for detecting characteristics of a plurality of reception frequency functions, and the detection means. An identification execution means (S15) for applying a feature to a classification model prepared in advance to identify a user's behavior mode is included. As a result, accurate behavior identification is possible.

  The behavior identification system according to the invention of claim 5 is dependent on claim 4, and the classification model is a decision tree.

  The behavior identification system according to the invention of claim 6 is dependent on claim 4 or 5, and further comprises a specifying means (S11) for specifying the moving direction of the user based on the reception frequency function created by the creating means, and the detecting means Performs the feature detection process with reference to the moving direction specified by the specifying means.

  The action identification system according to the invention of claim 7 is dependent on any one of claims 1 to 6, and the fixed pattern signal is an infrared signal.

  According to the present invention, the plurality of detection areas to which attention is paid by the plurality of detection units are different from each other with respect to at least one of size and position. For this reason, the plurality of reception frequency functions created by the creating unit depend on the user's behavior and are different from each other. A user's action mode is identified based on such a reception frequency function. That is, action identification with a simple configuration becomes possible.

  The above object, other objects, features and advantages of the present invention will become more apparent from the following detailed description of embodiments with reference to the drawings.

Referring to FIG. 1, an action identification system 10 of this embodiment includes an infrared receiver (hereinafter simply referred to as “receiver”) 12 installed on an upper portion of a gate GT and an infrared ray mounted on the back of a user H. And a transmitter (hereinafter simply referred to as “transmitter”) 14.
The receiver 12 includes two light receiving modules 12a and 12b and a main body 12c. The light receiving module 12a is provided on the side surface on the inlet side of the gate GT, and the light receiving module 12b is provided on the side surface on the outlet side of the gate GT. The light receiving modules 12a and 12b have detection areas Sa and Sb shown in FIG. 7, respectively. The detection areas Sa and Sb are different from each other in at least one of size and position. The light receiving modules 12a and 12b detect the infrared signal transmitted from the transmitter 14 while paying attention to the detection areas Sa and Sb, respectively, and output a detection signal of “0” or “1”.

The infrared signal transmitted from the transmitter 14 has an 8-bit fixed pattern (ID). As a result, the detection signal also shows an 8-bit fixed pattern. The transmission frequency of the ID is set to 50 Hz so that the reception frequency of the light receiving modules 12a and 12b is sufficiently different depending on the action of the user H. Moreover, CR2032 is used as a battery (not shown). The average battery driving time at this frequency is about 24 hours. This driving time is sufficient for tracking the daily behavior of the user H. Since the transmitter 14 is worn by the user H, the external size is reduced to 24 mm × 32 mm × 9.5 mm so as not to affect the behavior of the user H.
The main body 12c of the receiver 12 is configured as shown in FIG. The detection signal output from each of the light receiving modules 12a and 12b is given to the CPU 12f. The CPU 12f detects the two IDs by decoding the two given detection signals, respectively. The detected ID is registered in the table 12t shown in FIG. 3 together with the time stamp. Thus, time series data is created.

The CPU 12f creates two reception frequency functions indicating the ID reception frequency based on the time series data in the table 12t, and detects the passing direction of the user H from the two received reception frequency functions. The CPU 12f also extracts features of the two reception frequency functions with reference to the detected passing direction, and inputs the extracted features into a classification model such as a decision tree stored in the memory 12g. As a result, the action of the user H is identified. The identification result is transmitted to a server (not shown) via the wireless LAN 12h.
The reception frequency function is created in detail as follows. In the mathematical formulas described later, “A” is assigned to the parameter corresponding to the light receiving module 12a, and “B” is assigned to the parameter corresponding to the light receiving module 12b.

Referring to FIG. 4A, when the width of window W is Tw seconds and the shift width of window W once is ΔTw, reception indicating ID reception frequency per unit time by R∈ {A, B}. The frequency function h _R , id (t), (R = {A, B}) is expressed by Equation 1 and Equation 2.

According to the function f (t, R, id) shown in Equation 1, if the light receiving module with id received at time t is R, the calculation result indicates “1”; otherwise, the calculation result is “0”. Indicates. When the function f (t, R, id) is integrated over the time zone Tw with the time t as the midpoint time, the reception frequency function h _R , id (t), (R = {A, B}) is obtained. As a result, the reception frequency function shown in FIG. 4B is obtained for the light receiving module output shown in FIG.
For discretized data, Equation 3 is applied instead of Equation 2.

Hereinafter, when only a specific ID is described, or when the ID is clear from the context, the description of “id” is omitted. The reception frequency function corresponding to the light receiving module 12a is represented as h _A (t), and the reception frequency function corresponding to the light receiving module 12b is represented as h _B (t).

Assuming that the reception frequency functions h _A (t) and h _B (t) are drawn as shown in FIG. 5, the passing direction of the user H is detected as follows.

First, a time zone in which each of the reception frequency functions h _A (t) and h _B (t) has a value greater than “0” is defined by Equation 4.

The time zone in which the reception frequency function h _A (t) shows a value larger than “0” is defined by the start time t _A start and the end time t _A end. The time zone in which the reception frequency function h _B (t) indicates a value larger than “0” is defined by the start time t _B start and the end time t _B end. Further, the earlier time of the start times t _A start and t _B start is defined as Tstart, and the later time of the end times t _A end and t _B end is defined as Tend.

  Since the light receiving module on the side to which the user H approaches reacts first and the light receiving module after passing reacts with a delay, the passing direction of the user H is specified based on the above-described times Tstart and Tend. If the light receiving module that reacts first is “light receiving module before passing” and the light receiving module that reacts later is “light receiving module after passing”, the light receiving module that outputs the ID of the reception frequency function corresponding to Tstart is the light receiving module before passing. The light receiving module that outputs the ID of the reception frequency function corresponding to Tend is specified as the light receiving module after passing. Thereby, the passing direction of the user H can be detected.

When the passing direction is detected, the characteristics of the reception frequency functions h _A (t) and h _B (t) are extracted as follows.
First, a time zone Ttot in which one of the reception frequency functions h _A (t) and h _B (t) has a value larger than “0” is obtained by Equation 5.

Next, the reception frequency function h _R (t), (R = {A, B}) is converted so that Tstart becomes time 0. The variable reception frequency function is expressed as h _R (τ), (τ = t−Tstart). Integrated value [Phi _R per unit time received frequency function h _R (τ) is determined by the number 6, the value phi _R to the integrated value [Phi _R normalized is determined by the number 7.

The time centroid Tg _R , R = {A, B} of the reception frequency function h _R (τ) is obtained by Equation 8. Further, the time difference ΔTg of the time centroid is obtained by Equation 9.

In Equation 9, Tg_atfer is the time centroid of the reception frequency function corresponding to the light-receiving module after passage, and Tg_before is the time centroid of the reception frequency function corresponding to the light-receiving module before passage. The normalized time difference δTg is obtained by Equation 10.

  The decision tree stored in the memory 14g is configured as shown in FIG. The characteristics Ttot, φafter, φbefore, Tg_atfer, Tg_before, ΔTg, and δTg obtained as described above are input to this decision tree. As a result, whether the action of the user H is “walk”, “run”, “through”, “turn”, or “wheelchair” is identified. Φafter is a normalized integrated value of the reception frequency function corresponding to the post-passing light receiving module, and φbefore is a normalized integrated value of the reception frequency function corresponding to the pre-passing light receiving module.

When the user H takes action I shown in FIG. 7, the reception frequency function changes as shown in FIG. At this time, the behavior mode of the user H is identified as “walk”. When the user H takes action II shown in FIG. 7, the reception frequency function changes as shown in FIG. 8B. At this time, the behavior mode of the user H is identified as “run”. When the user H takes action III shown in FIG. 7, the reception frequency function changes as shown in FIG. At this time, the behavior mode of the user H is identified as “through”.
When the user H takes action IV shown in FIG. 7, the reception frequency function changes as shown in FIG. At this time, the behavior mode of the user H is identified as “turn”. When the user H takes action V shown in FIG. 7, the reception frequency function changes as shown in FIG. At this time, the behavior mode of the user H is identified as “wheelchair”.

  For actions I, II and V, the value of the reception frequency function corresponding to the light receiving module 12a is smaller than the value of the reception frequency function corresponding to the light receiving module 12b. This is because the transmitter 14 is mounted on the back of the user H, and the infrared signal detection accuracy differs between the front and the rear of the user H. As a result, a clear difference appears in the reception frequency function, and the accuracy of action identification increases.

The above-described decision tree is created by letting the user H perform pre-training behavior to acquire time series data and assign action labels to features extracted from the reception frequency function based on this time series data. The
Specifically, the CPU 12f shown in FIG. 2 executes processing according to the flowchart shown in FIG. First, in step S1, a plurality of settings including a time stamp are initialized. In step S3, it is determined whether or not an infrared signal is received by the light receiving module 12a or 12b. If YES, the processes in and after step S5 are executed.

  In step S5, the received infrared signal, that is, the detection signal is decoded to detect the ID, and the detected ID is written in the table 12t together with the time stamp. Thus, time series data is created. The created time series data is sorted for each ID in step S7.

In step S9, the calculation according to the above formulas 1 and 2 is executed based on the sorted time series data, and two reception frequency functions are obtained. In step S11, the passing direction of the user H is detected using the above-described equation 4, and in step S13, the calculation according to the above-described equations 5 to 10 is executed to extract the characteristics of the reception frequency function.
In step S15, the extracted feature is input to the decision tree to identify the user's action. In step S17, the identification result is transmitted to the server through the wireless LAN 12h. In step S19, it is determined whether or not an end instruction has been issued. If NO, the process returns to step S3. If YES, the process ends.

  As can be seen from the above description, the transmitter 14 that transmits an infrared signal having a fixed pattern (ID) is mounted on the back of the user H who acts arbitrarily. On the other hand, the light receiving modules 12a and 12b detect the infrared signal transmitted from the transmitter 14 by paying attention to two detection areas Sa and Sb that are different from each other in at least one of size and position. The CPU 12f creates a reception frequency function indicating the reception frequency per unit time of the detection signal by each of the light receiving modules 12a and 12b (S9). The behavior mode of the user H is identified by applying the characteristics of the two generated reception frequency functions to a classification model such as a decision tree (S15).

  The two detection areas Sa and Sb noted by the light receiving modules 12a and 12b are different from each other with respect to at least one of size and position. For this reason, the two reception frequency functions created by the CPU 12f depend on the action of the user H and are different from each other. The action mode or the passing direction of the user H is identified or detected based on such a reception frequency function. That is, action identification or passage direction detection can be performed with a simple configuration.

  In this embodiment, infrared signals are received by two light receiving modules, but the number of light receiving modules may be three or more. In this embodiment, an infrared signal is used. However, the present invention is not limited to this as long as it is wireless. Furthermore, in this embodiment, a decision tree is used as a classification model for behavioral modes. Instead, a k-NN (Nearest Neighbor) method, an SVM (Support Vector Machine) method, a DP (Dynamic Programming) method, an HMM (Hidden Markov Model) method may be applied.

  In this embodiment, the behavior mode of the user H is specified from “walk”, “run”, “through”, “turn”, and “wheelchair”. “Skip”, vehicles other than wheelchairs are also conceivable. Furthermore, in this embodiment, the transmitter is mounted on the back of the user's head, but the mounting position is not limited to the back of the head. In this embodiment, a receiver is provided at the gate. However, the receiver may be provided at the top of the cross or T-junction, the ceiling, or the like.

It is an illustration figure which shows one Example of this invention. It is a block diagram which shows an example of a structure of the main body of the infrared transmitter applied to the FIG. 1 Example. It is an illustration figure which shows an example of the table applied to the FIG. 2 Example. (A) is a wave form diagram which shows an example of the change of the infrared signal output from a light reception module, (B) is a wave form diagram which shows an example of the reception frequency function calculated | required based on the infrared signal shown to (A). is there. It is a graph which shows another example of a reception frequency function. It is an illustration figure which shows an example of a structure of a decision tree. It is an illustration figure which shows an example of a user's action. (A) is a graph showing an example of a reception frequency function corresponding to action I, (B) is a graph showing an example of a reception frequency function corresponding to action II, and (C) is a reception corresponding to action III. It is a graph which shows an example of a frequency function, (D) is a graph which shows an example of the reception frequency function corresponding to action IV, (E) is a graph which shows an example of the reception frequency function corresponding to action V. It is a flowchart which shows an example of operation | movement of CPU.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 ... Action identification system 12 ... Infrared receiver 14 ... Infrared transmitter 12a, 12b ... Light receiving module 12f ... CPU
12h ... Wireless LAN

Claims (7)

Transmitting means for transmitting a fixed pattern signal held by a user;
A plurality of detection means for detecting fixed pattern signals sent from the transmission means by paying attention to a plurality of detection areas having different sizes and positions from each other,
Creation means for creating a reception frequency function indicating a reception frequency per unit time of the fixed pattern signal detected by each of the plurality of detection means, and the user based on the plurality of reception frequency functions created by the creation means A behavior identification system comprising identification means for identifying the behavior mode. The transmitting means is worn on the user's body;
The behavior identification system according to claim 1, wherein the plurality of detection units are attached to different positions of the fixed object. The fixed object is a gate;
3. The behavior identification system according to claim 2, wherein the plurality of detection units include a first detection unit provided on a side surface on the entrance side of the gate and a second detection unit provided on a side surface on the exit side of the gate.   The identification unit is a detection unit that detects features of the plurality of reception frequency functions, and an identification execution unit that applies the features detected by the detection unit to a classification model prepared in advance to identify the behavior mode of the user. The action identification system according to claim 1, comprising:   The behavior identification system according to claim 4, wherein the classification model is a decision tree. Further comprising specifying means for specifying the moving direction of the user based on the reception frequency function created by the creating means;
The behavior detection system according to claim 4, wherein the detection unit executes a feature detection process with reference to the moving direction specified by the specification unit.   The behavior identification system according to claim 1, wherein the fixed pattern signal is an infrared signal.

Images