Behavior prediction device, behavior prediction method, and behavior prediction program

Abstract

Provided is an action prediction device for predicting a future action of a specific person, the action prediction device comprising: an information extraction unit that extracts, on the basis of a past action series representing a past action history of the specific person as a time series, start time probability distribution data representing a probability distribution of a start time for each type of action, action transition probability data representing a probability of transition from one action to another action for each combination of types of action, and action period data relating to an action period for each type of action; and an action prediction unit that predicts the future action of the specific person on the basis of the start time probability distribution data, the action transition probability data, the action period data, and a current action of the specific person.

Description

[Technical Field]

[0001] The disclosed technology relates to a behavior prediction device, a behavior prediction method, and a behavior prediction program. [Background technology]

[0002] Conventionally, there are known techniques for predicting a person's future behavior. For example, Patent Document 1 describes a technique for predicting a user's behavior based on the user's place of stay, the time of stay, and the occurrence probability of each past behavior pattern. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2010-250759 Summary of the Invention [Problem to be solved by the invention]

[0004] However, the technology described in Patent Document 1 uses a behavior information pattern that indicates the start time of the behavior, the end time of the behavior, and the transition behavior of the place of stay as factors for determining the next behavior. However, the factors described in Patent Document 1 alone are not sufficient factors for predicting future behavior. Therefore, there is room for improvement in the accuracy of predicting a person's behavior.

[0005] The disclosed technology has been made in consideration of the above points, and aims to provide a behavior prediction device, a behavior prediction method, and a behavior prediction program that can predict the future behavior of a specific person with higher accuracy. [Means for solving the problem]

[0006] A first aspect of the present disclosure is a behavior prediction device that predicts future behavior of a specific person, the behavior prediction device comprising: an information extraction unit that extracts, based on a past behavior sequence that represents the specific person's past behavior history in a time series, start time probability distribution data that represents a probability distribution of start times for each type of behavior, behavior transition probability data that represents the probability of transitioning from one behavior to another for each combination of behavior types, and behavior time data related to the behavior time for each of the behavior types; and a behavior prediction unit that predicts the future behavior of the specific person based on the start time probability distribution data, the behavior transition probability data, and the behavior time data, and the specific person's current behavior.

[0007] A second aspect of the present disclosure is a behavior prediction method for predicting the future behavior of a specific person, wherein an information extraction unit extracts start time probability distribution data representing the probability distribution of start times for each type of behavior, behavior transition probability data representing the probability of transitioning from one behavior to another for each combination of behavior types, and behavior time data relating to the behavior time for each type of behavior, based on a past behavior series that represents the past behavior history of the specific person in a chronological order, and a behavior prediction unit predicts the future behavior of the specific person based on the start time probability distribution data, the behavior transition probability data, and the behavior time data, and the current behavior of the specific person.

[0008] A third aspect of the present disclosure is a behavior prediction program for causing a computer to function as each unit of the behavior prediction device of the first aspect. [Effects of the Invention]

[0009] According to the disclosed technology, it is possible to predict the future behavior of a specific person with higher accuracy. [Brief explanation of the drawings]

[0010] [Figure 1] FIG. 1 is a configuration diagram illustrating an example of a configuration of a behavior prediction system according to an embodiment. [Figure 2]FIG. 2 is a diagram illustrating an example of a hardware configuration of a behavior prediction device according to an embodiment. [Figure 3] 1 is a block diagram illustrating an example of a functional configuration of a behavior prediction device according to an embodiment. [Figure 4] FIG. 10 is a diagram showing an example of transition of a specific person's behavior over two weeks. [Figure 5A] FIG. 10 is a diagram illustrating an example of behavior transition probability data. [Figure 5B] FIG. 10 is a diagram illustrating an example of elapsed time probability distribution data. [Figure 5C] FIG. 10 is a diagram illustrating an example of start time probability distribution data. [Figure 5D] FIG. 10 is a diagram illustrating an example of end time probability distribution data. [Figure 5E] FIG. 10 is a diagram illustrating an example of behavior time probability distribution data. [Figure 6] 10 is a diagram for explaining the operation of a occurring behavior / behavior start time prediction unit. FIG. [Figure 7] 10 is a diagram for explaining the operation of a behavior end time prediction unit. FIG. [Figure 8] 10 is a flowchart illustrating an example of information extraction processing in the behavior prediction device of the embodiment. [Figure 9] 10 is a flowchart illustrating an example of a behavior prediction process in the behavior prediction device of the embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0011] An example of an embodiment of the disclosed technology will be described below with reference to the drawings. Note that the same or equivalent components and parts in each drawing are given the same reference numerals. Also, the dimensional proportions in the drawings are exaggerated for the convenience of explanation and may differ from the actual proportions.

[0012] First, an example of the configuration of a behavior prediction system 1 according to the technology of the present embodiment will be described. As shown in Fig. 1, the behavior prediction system 1 according to the present embodiment includes a behavior prediction device 10 and a sensor group 12. The behavior prediction device 10 and the sensor group 12 are connected via a network 9 by wired communication or wireless communication.

[0013] The sensor group 12 includes multiple types of sensors used to detect events that change due to the behavior of a person. Specific types of sensors vary depending on the events that change due to the behavior of the person being detected, but examples include sensors that detect a person's vital signs, a place where the person is staying, temperature, humidity, illuminance, sound volume, a person's sleep state, a person's wakefulness state, and electricity consumption. The detection results of each sensor included in the sensor group 12 are output to the behavior prediction device 10 via the network 9.

[0014] The behavior prediction device 10 is a device that predicts the future behavior of a specific person. The behavior prediction device 10 of this embodiment estimates the current behavior of a specific person based on the detection results of events that change due to the behavior of the specific person, which are input from a sensor group 12. The behavior prediction device 10 also predicts the future behavior of the specific person based on statistical information extracted from a past behavior sequence that represents the specific person's past behavior history in a time series, and the specific person's current behavior. Note that the "current behavior" refers to the behavior at the start time of the prediction period, which is the starting point for predicting the specific person's future behavior. Therefore, the "current behavior" may differ from the behavior at the time when the behavior prediction process (see FIG. 9), which will be described later, is executed.

[0015] Fig. 2 is a configuration diagram showing an example of the hardware configuration of the behavior prediction device 10 of this embodiment. As shown in Fig. 2, the behavior prediction device 10 includes a CPU (Central Processing Unit) 21, a ROM (Read Only Memory) 22, a RAM (Random Access Memory) 23, a storage 24, a display unit 30, and a communication I / F (Interface) 32. Each component is connected to each other so as to be able to communicate with each other via a bus 39 such as a system bus or a control bus.

[0016] The CPU 21 is a central processing unit, and executes various programs such as the behavior prediction program 35 stored in the storage 24, and controls each part.

[0017] The ROM 22 stores various programs and various data to be executed by the CPU 21. The RAM 23 temporarily stores the programs or data as a work area when the CPU 21 executes the various programs. That is, the CPU 21 reads the programs from the storage 24 and executes the programs using the RAM 23 as a work area.

[0018] The storage 24 of this embodiment stores a behavior prediction program 25A and a statistical information extraction program 25B. Each of the behavior prediction program 25A and the statistical information extraction program 25B may be a single program or a group of programs configured from multiple programs or modules. The storage 24 is configured with an HDD (Hard Disk Drive) or an SSD (Solid State Drive). The storage 24 also stores various programs including an operating system and various data (neither of which is shown). The storage 24 also stores a behavior estimation model 26 used to estimate a person's behavior.

[0019] The display unit 30 displays various information such as information about the future behavior of a specific person, which is the prediction result, etc. The display unit 30 is not particularly limited, and various displays may be used.

[0020] The communication I / F 32 is an interface for communicating with the sensor group 12 via the network 9, and uses standards such as Ethernet (registered trademark), FDDI, and Wi-Fi (registered trademark).

[0021] Next, a functional configuration of the behavior prediction device 10 will be described. As shown in Fig. 3, the behavior prediction device 10 includes a feature extraction unit 40, a behavior estimation unit 42, an information extraction unit 44, a behavior prediction unit 46, and a display control unit 52. When the CPU 21 executes the statistical information extraction program 25B stored in the storage 24, the CPU 21 functions as the feature extraction unit 40, the behavior estimation unit 42, and the information extraction unit 44. When the CPU 21 executes the behavior prediction program 25A stored in the storage 24, the CPU 21 functions as the feature extraction unit 40, the behavior estimation unit 42, the behavior prediction unit 46, and the display control unit 52.

[0022] The feature extraction unit 40 has a function of extracting feature amounts related to events that change due to the behavior of a person from the detection results input from the sensor group 12. The extracted feature amounts depend on the behavior of the person, and examples include the person's vital signs, the place where the person is staying, temperature, humidity, illuminance, sound volume, the person's sleeping state, the person's waking state, and electricity consumption. The feature extraction unit 40 outputs the extracted feature amounts to the behavior estimation unit 42.

[0023] The behavior estimation unit 42 has a function of deriving the type of behavior (past behavior sequence 60 or current behavior 62) of a specific person based on the feature input from the feature extraction unit 40 using the behavior estimation model 26.

[0024] The activity estimation model 26 is a trained model that receives the features extracted by the feature extraction unit 40 as input and outputs the type of activity. Examples of the type of activity include, but are not limited to, waking up, getting ready, cooking, eating, work, entertainment, and sleeping.

[0025] As an example, the behavior estimation model 26 of this embodiment is a model based on a decision tree. The behavior estimation model 26 is obtained, for example, by training a learning model in the following learning phases. In the learning phase, the behavior estimation model 26 is trained by being given learning data, also called training data or teacher data. The training data is a set of feature quantities extracted by the feature extraction unit 40 and correct behavior types. In the learning phase, the feature quantities extracted by the feature extraction unit 40 are vectorized and input to the behavior estimation model 26. The behavior estimation model 26 outputs the behavior type as an estimation result for the feature quantities extracted by the feature extraction unit 40. The behavior estimation model 26 is trained by optimizing each branch of the decision tree based on the behavior type that is the estimation result and the behavior type that is the correct answer.

[0026] As in the present embodiment, the behavior prediction device 10 may acquire the behavior estimation model 26 that has been trained by an external device and store it in the storage 24. Also, unlike the present embodiment, the behavior prediction device 10 may learn the behavior estimation model 26. Also, unlike the present embodiment, the behavior prediction device 10 may not store the behavior estimation model 26, and the behavior estimation unit 42 may use the behavior estimation model 26 stored in an external device.

[0027] The behavior estimation model 26 may be a model specialized for each person, or may be a general-purpose model. Specifically, the behavior estimation model 26 may be specialized for the behavior of a specific person by using, as training data, only feature amounts extracted by the feature amount extraction unit 40 from the detection results of the sensor group 12 related to the specific person. Alternatively, the behavior estimation model 26 may be specialized for the behavior of a general person by using, as training data, feature amounts extracted by the feature amount extraction unit 40 from the detection results of the sensor group 12 related to multiple people.

[0028] The behavior estimation unit 42 inputs the feature amounts received from the feature amount extraction unit 40 into the behavior estimation model 26, and acquires the type of behavior of the person output from the behavior estimation model 26. When generating statistical information 28 related to the behavior of a specific person, the behavior estimation unit 42 outputs a past behavior sequence 60, which represents the type of behavior of the specific person in a time series, to the information extraction unit 44. FIG. 4 shows, as an example, transitions in the behavior of a specific person over two weeks. In this case, the behavior estimation unit 42 outputs, to the information extraction unit 44, the past behavior sequence 60, which represents the type of behavior corresponding to the transition example shown in FIG. 4 in a time series. In this way, the past behavior sequence 60 represents the past behavior history of the specific person in a time series.

[0029] When generating statistical information 28 regarding the behavior of a specific person, a past behavior sequence 60 is input to the information extraction unit 44 from the behavior estimation unit 42. The information extraction unit 44 has a function of extracting statistical information 28 from the past behavior sequence 60. In other words, the information extraction unit 44 has a function of generating statistical information 28 using the past behavior sequence 60 as learning data. As shown in FIG. 3 , the statistical information 28 includes behavior transition probability data 28A, elapsed time probability distribution data 28B, start time probability distribution data 28C, end time probability distribution data 28D, and behavior time probability distribution data 28E.

[0030] The behavior transition probability data 28A is data that represents the probability (transition probability P(At-1→At) of transitioning from one behavior (behavior At-1) to another behavior (behavior At) for each combination of behavior types. FIG. 5A shows an example of the behavior transition probability data 28A. As shown in FIG. 5A, the behavior transition probability data 28A of this embodiment represents transition behavior from one behavior to the next behavior as a Markov model. Note that FIG. 5A shows only some of the behavior types and transition behaviors, and the actual behavior transition probability data 28A represents transition behaviors for all combinations of behavior types that are the subject of estimation.

[0031] The elapsed time probability distribution data 28B is data that represents the probability distribution of the elapsed time from the start of one action (action At-1) to the start of the other action (action At) for each combination of action types. FIG. 5B shows an example of the elapsed time probability distribution data 28B. The action time for performing a certain action (action At-1) may differ depending on the type of the next action (action At). For example, as shown in FIG. 5B, when getting ready is performed after getting up, the action time for getting up may differ from the action time for having breakfast when having breakfast after getting up. Specifically, the elapsed time from starting to get up until starting to get ready, which is the next action, may differ from the elapsed time from starting to get up until starting to have breakfast, which is the next action.

[0032] 5B, for example, when getting ready and then getting up, the time taken to get ready may differ from the time taken to have breakfast when getting ready and then having breakfast. Specifically, the time elapsed from starting to get ready to start getting up, which is the next action, may differ from the time taken from starting to get ready to start having breakfast, which is the next action.

[0033] Therefore, the information extraction unit 44 of this embodiment extracts, for each combination of behavior types, data representing the probability distribution of the elapsed time from the start of one behavior (behavior At-1) to the start of the other behavior (behavior At), as elapsed time probability distribution data 28B. In this way, the elapsed time probability distribution data 28B is data indicating the tendency of behavior time according to the type of the next behavior (behavior At).

[0034] The start time probability distribution data 28C is data that represents the probability distribution of the start time of each behavior (behavior At). Fig. 5C shows an example of the start time probability distribution data 28C. Fig. 5C also shows the probability distribution of the start time when the type of behavior is waking up and the probability distribution of the start time when the type of behavior is having breakfast.

[0035] The end time probability distribution data 28D is data that represents the probability distribution of the end time for each type of activity. Fig. 5D shows an example of the end time probability distribution data 28D. Fig. 5D also shows the probability distribution of the end time when the type of activity is waking up and the probability distribution of the end time when the type of activity is breakfast.

[0036] The activity time probability distribution data 28E is data that represents the probability distribution of the activity time during which an activity is performed for each activity type. Fig. 5E shows an example of the activity time probability distribution data 28E. Fig. 5E also shows the probability distribution of the activity time when the activity type is waking up and the probability distribution of the activity time when the activity type is breakfast.

[0037] The information extraction unit 44 stores the extracted statistical information 28 in the storage 24 .

[0038] On the other hand, when predicting the future behavior of a specific person, the behavior estimation unit 42 outputs a current behavior 62 indicating the type of current behavior of the specific person to the behavior prediction unit 46. As shown in FIG. 3 , the behavior prediction unit 46 includes an occurring behavior / behavior start time prediction unit 48 and an behavior end time prediction unit 50.

[0039] The occurring behavior / behavior start time prediction unit 48 has a function of predicting the type of next behavior (behavior At) and the start time of that behavior At based on the behavior transition probability data 28A, the elapsed time probability distribution data 28B, and the start time probability distribution data 28C. Fig. 6 shows an example of predicting the type of behavior (behavior At) next to the behavior of going to bed (behavior At-1) and the start time of that behavior At. The occurring behavior / behavior start time prediction unit 48 derives a score S (score) for each type of behavior, taking into account both the start time of the behavior and the elapsed time until the next behavior, in timebox units obtained by dividing time into predetermined units (30 minutes in Fig. 6), using the following formula (1): S={Pmax(start|At)}+Pmax(elapsed|At-1→At)+P(At-1→At) ···(1)

[0040] where Pmax(start|At) is the maximum probability within the timebox in the start time probability distribution, Pmax(elapsed|At-1→At) is the maximum probability within the timebox in the elapsed time probability distribution, and P(At-1→At) is the transition probability from action At-1 to action At.

[0041] 6, when the type of activity is getting up, the score S1 of the timebox from 5:00 to 5:30 is the highest, so the score S1 becomes the score S used to estimate candidates for the next activity At, and the timebox from 5:00 to 5:30 is selected. When the type of activity is getting ready, the score S3 of the timebox from 6:00 to 6:30 is the highest, so the score S3 becomes the score S used to estimate candidates for the next activity At, and the timebox from 6:00 to 6:30 is selected.

[0042] Furthermore, the occurring behavior / behavior start time prediction unit 48 derives start time candidates for each type of behavior in the timebox selected by the score S using the following equation (2). Start time (candidate for action At) = {start of Pmax(start|At) + (start time of At-1 + time of Pmax(elapsed|At-1→At)} ÷ 2 (2)

[0043] Here, the start of Pmax(start|At) is the start time with the maximum probability within the timebox in the start time probability distribution, and the time of Pmax(elapsed|At-1→At) is the elapsed time with the maximum probability within the timebox in the elapsed time probability distribution.

[0044] Furthermore, the occurring behavior / behavior start time prediction unit 48 compares the scores S of each behavior type and estimates the behavior type with the maximum score S as the next behavior At. For example, in the above example, since the score S1 of getting up is greater than the score S3 of getting ready, getting up is estimated as the next behavior At, and the start time of getting up obtained by the above formula (2) is predicted as the start time of the next behavior At.

[0045] The occurring behavior / behavior start time prediction unit 48 outputs the predicted next behavior At and its start time to the behavior end time prediction unit 50 .

[0046] The behavior end time prediction unit 50 has a function of predicting the end time of the next behavior At estimated by the behavior occurrence / behavior start time prediction unit 48, based on the end time probability distribution data 28D and the behavior duration probability distribution data 28E. FIG. 7 shows an example of predicting the end time when the next behavior At is waking up. The behavior occurrence / behavior start time prediction unit 48 derives a score S (score) for each behavior, taking into account both the behavior end time and the behavior duration, for each timebox unit, which is a predetermined unit of time (30 minutes in FIG. 7), using the following formula (3): S={Pmax(end|At)}+Pmax(time|At) ···(3)

[0047] Here, Pmax(end|At) is the maximum probability within the timebox in the end time probability distribution, and Pmax(time|At) is the maximum probability within the timebox in the activity time probability distribution.

[0048] In the example shown in FIG. 7, when the type of behavior is waking up, the score S2 is the highest, so 5:30 to 6:00, which is the timebox corresponding to waking up S2, is selected as a candidate for the end time.

[0049] Furthermore, the behavior end time prediction unit 50 derives a candidate end time for the estimated next behavior At in the selected timebox using the following equation (4). End time (candidate for action At) = {end of Pmax(end|At) + (start time of At + time of Pmax(time|At)} ÷ 2 (4)

[0050] Here, the end in Pmax(end|At) is the end time with the maximum probability within the timebox in the end time probability distribution, and the time in Pmax(time|At) is the behavior time with the maximum probability within the timebox in the behavior time probability distribution.

[0051] The behavior end time prediction unit 50 outputs the predicted end time of the next behavior At.

[0052] The behavior prediction unit 46 outputs the next behavior At and its start time predicted by the occurring behavior / behavior start time prediction unit 48, and the end time of the next behavior At predicted by the behavior end time prediction unit 50 to the display control unit 52.

[0053] The display control unit 52 has a function of displaying, in chronological order, the types of actions and their start and end times regarding the future actions of a specific person estimated by the action estimation unit 42 on the display unit 30. Note that instead of or in addition to displaying, the types of actions and their start and end times may be stored in the storage 24.

[0054] Next, the operation of the behavior prediction device 10 of this embodiment will be described.

[0055] Fig. 8 shows a flowchart of an example of statistical information extraction processing executed by the behavior prediction device 10 of this embodiment. The behavior prediction device 10 executes the statistical information extraction processing shown in Fig. 8 by executing the statistical information extraction program 25B stored in the storage 24. Note that the statistical information extraction processing shown in Fig. 8 is executed at a predetermined timing, such as when an execution instruction is received from a user.

[0056] In step S10 of FIG. 8, the information extraction unit 44 acquires the past behavior sequence 60 as described above.

[0057] In the next step S12, the information extraction unit 44 extracts behavior transition probability data 28A, elapsed time probability distribution data 28B, start time probability distribution data 28C, end time probability distribution data 28D, and behavior duration probability distribution data 28E as statistical information 28, as described above. Note that the method of extracting each of the behavior transition probability data 28A, elapsed time probability distribution data 28B, start time probability distribution data 28C, end time probability distribution data 28D, and behavior duration probability distribution data 28E is not particularly limited, and known methods can be used. When the processing of step S12 is completed, the statistical information extraction process shown in FIG. 8 is completed.

[0058] After the statistical information 28 is extracted in this manner, the behavior prediction device 10 executes the behavior prediction process shown in FIG. 9. The behavior prediction process shown in FIG. 9 is a process for predicting behaviors A0, A1, A2, A3, ... of a specific person from the present time until the end time of the prediction period. FIG. 9 shows a flowchart of an example of the behavior prediction process executed by the behavior prediction device 10 of this embodiment. The behavior prediction device 10 executes the behavior prediction process shown in FIG. 9 by executing a behavior prediction program 25A stored in the storage 24. Note that the behavior prediction process shown in FIG. 9 may be executed immediately after the statistical information extraction process shown in FIG. 8 is completed, or may be executed when an execution instruction is received from a user after the statistical information extraction process, or may be executed at a predetermined timing.

[0059] In step S100 of FIG. 9, the feature extraction unit 40 sets the variable t to zero (t=0).

[0060] In the next step S102, the feature extraction unit 40 acquires the detection results of the sensor group 12 as described above.

[0061] In the next step S104, the feature extraction unit 40 extracts feature amounts relating to events that change due to the behavior of a person from the detection results input from the sensor group 12, as described above.

[0062] In the next step S106, the behavior estimation unit 42 estimates the current behavior of the specific person. As described above, the behavior estimation unit 42 inputs the features extracted by the feature extraction unit 40 to the behavior estimation model 26, and acquires the type of behavior output from the behavior estimation model 26 as the current behavior 62.

[0063] In the next step S108, the behavior end time prediction unit 50 of the behavior prediction unit 46 predicts the behavior end time of the current behavior At-1. As described above, since t=0, the current behavior At-1=A0 here. As described above, the behavior end time prediction unit 50 of this embodiment selects the timebox that maximizes the score S for the end time of the current behavior At-1. Then, the behavior end time prediction unit 50 predicts the end time of the current behavior At-1 using the following equation (5), in which "next behavior At" in the above equation (4) is changed to the current behavior "At-1." End time (candidate for current action At-1) = {end of Pmax(end|At-1) + (start time of At-1 + time of Pmax(time|At-1)} ÷ 2 (5)

[0064] In the next step S110, the occurrence behavior / behavior start time prediction unit 48 determines whether the end time derived in step S108 has reached the end time of the prediction period. In this embodiment, a prediction period is set for predicting future behavior of a specific person. For example, when predicting behavior up to 24 hours from now, the prediction period is 24 hours. The prediction period may be set in advance in the behavior prediction device 10, or may be set by the user who performs the prediction. Until the end time of the prediction period is reached, the determination in step S110 remains negative, and the process proceeds to step S112.

[0065] In step S112, the occurring behavior / behavior start time prediction unit 48 predicts the type and start time of the next behavior At. As described above, the occurring behavior / behavior start time prediction unit 48 predicts the type and start time of the next behavior At using the above equations (1) and (2) based on the behavior transition probability data 28A, the elapsed time probability distribution data 28B, and the start time probability distribution data 28C.

[0066] In the next step S114, the behavior end time prediction unit 50 predicts the end time of the next behavior At predicted in the above step S112. As described above, the behavior end time prediction unit 50 predicts the end time of the next behavior At using the above equations (3) and (4) based on the end time probability distribution data 28D and the behavior duration probability distribution data 28E.

[0067] In the next step S116, the occurrence behavior / behavior start time prediction unit 48 adds 1 to the variable t (t=t+1) in preparation for predicting the next behavior, then returns to step S11, and repeats the processes of steps S110 to S116 until the end time of the prediction period is reached. This sequentially makes predictions for behaviors A2, A3, ...

[0068] On the other hand, if the end time of the prediction period has been reached in step S110, the determination is affirmative, and the process proceeds to step S118.

[0069] In step S118, the display control unit 52 displays the prediction result on the display unit 30. As described above, the display control unit 52 displays, in chronological order, the types of behaviors and their start times and end times regarding the future behaviors of a specific person estimated by the behavior estimation unit 42 on the display unit 30. When the processing of step S118 ends, the behavior prediction processing shown in FIG. 9 ends.

[0070] As described above, the behavior prediction device 10 of this embodiment is a behavior prediction device that predicts the future behavior of a specific person, and includes an information extraction unit 44 and a behavior prediction unit 46.

[0071] Based on a past behavior sequence 60 that represents the past behavior history of a specific person in chronological order, the information extraction unit 44 extracts start time probability distribution data 28C that represents the probability distribution of the start time for each behavior type, end time probability distribution data 28D that represents the probability distribution of the end time for each behavior type, behavior transition probability data 28A that represents the probability of transitioning from one behavior to the other for each combination of behavior types, and behavior time probability distribution data 28E regarding the behavior time for each behavior type.The behavior prediction unit 46 predicts the future behavior of the specific person based on the start time probability distribution data 28C, end time probability distribution data 28D, behavior transition probability data 28A, and behavior time probability distribution data 28E, and the specific person's current behavior 62.

[0072] Furthermore, the information extraction unit 44 of this embodiment further extracts elapsed time probability distribution data 28B, and the behavior prediction unit 46 predicts the future behavior of a specific person based on the elapsed time probability distribution data 28B as well.

[0073] As described above, the behavior prediction device 10 of this embodiment predicts the future behavior of a specific person by taking into account the behavior time for each behavior type. This allows for the trend of behavior time to be taken into account, making it possible to deal with time shifts such as a shift in the start time and end time while the behavior time remains the same, or a shift in the time while the combination of behavior types that transition to the next behavior remains the same. Furthermore, for behavior types that have periodic occurrence times such as work or lunch, it is possible to deal with specific start times and specific end times.

[0074] Therefore, the behavior prediction device 10 of this embodiment can predict the future behavior of a specific person with higher accuracy.

[0075] Although the behavior prediction device 10 of the present embodiment has been described as predicting the future behavior of a specific person using the end time probability distribution data 28D as well, it may also be configured to predict the future behavior of a specific person without using the end time probability distribution data 28D. For example, the behavior prediction device 10 may predict the end time from the start time probability distribution data 28C and the behavior time probability distribution data 28E, and predict the future behavior of the specific person using the predicted end time.

[0076] In addition, in the present embodiment, a configuration has been described in which the behavior of a specific person is estimated using the behavior estimation model 26 based on the quantity of features extracted from the detection results of the sensor group 12, but the method of estimating the behavior of a specific person is not limited to this configuration. For example, a configuration may be adopted in which the specific person himself or a user observing the behavior of the specific person inputs the type of behavior of the specific person to the behavior prediction device 10. Also, for example, a configuration may be adopted in which the behavior prediction device 10 estimates the behavior of a specific person from the start time probability distribution data 28C and the time.

[0077] Furthermore, the various processes executed by the CPU after reading software (programs) in the above embodiments may be executed by various processors other than the CPU. Examples of processors in this case include dedicated electrical circuits, such as programmable logic devices (PLDs) (such as field-programmable gate arrays (FPGAs)) whose circuit configuration can be changed after manufacture, and application-specific integrated circuits (ASICs) that are processors with circuit configurations specifically designed to execute specific processes. The behavior prediction process may be executed by one of these various processors, or by a combination of two or more processors of the same or different types (e.g., multiple FPGAs, or a combination of a CPU and an FPGA). The hardware structure of these various processors is, more specifically, an electrical circuit that combines circuit elements such as semiconductor devices.

[0078] In addition, in the above-described embodiments, the behavior prediction program 25A and the statistical information extraction program 25B are each pre-stored (installed) in the storage 24, but this is not limiting. The behavior prediction program 25A and the statistical information extraction program 25B may each be provided in a form stored in a non-transitory storage medium such as a CD-ROM (Compact Disk Read Only Memory), a DVD-ROM (Digital Versatile Disk Read Only Memory), or a USB (Universal Serial Bus) memory. Furthermore, the behavior prediction program 25A and the statistical information extraction program 25B may each be downloaded from an external device via a network.

[0079] The following additional notes are provided regarding the above-described embodiments.

[0080] (Additional note 1) Memory and at least one processor coupled to said memory; Including, The processor: Based on a past behavior sequence that represents the past behavior history of a specific person in a time series, start time probability distribution data that represents a probability distribution of start times for each type of behavior, behavior transition probability data that represents the probability of transitioning from one behavior to another for each combination of behavior types, and behavior time data related to behavior time for each type of behavior are extracted; predicting a future behavior of the specific person based on the start time probability distribution data, the behavior transition probability data, the behavior time data, and the current behavior of the specific person; The behavior prediction device is configured as follows.

[0081] (Additional note 2) A non-transitory storage medium storing a program executable by a computer to perform a behavior prediction process, The behavior prediction process includes: Based on a past behavior sequence that represents the past behavior history of a specific person in a time series, start time probability distribution data that represents a probability distribution of start times for each type of behavior, behavior transition probability data that represents the probability of transitioning from one behavior to another for each combination of behavior types, and behavior time data related to the behavior time for each of the behavior types are extracted; predicting a future behavior of the specific person based on the start time probability distribution data, the behavior transition probability data, the behavior time data, and the current behavior of the specific person; Non-transitory storage medium. [Explanation of symbols]

[0082] 10 Behavioral prediction device 12 Sensors 21 CPU 22 ROM 23 RAM 24 Storage 25A Behavioral prediction program, 25B Statistical information extraction program 26 Behavioral estimation model 28 Statistical information, 28A Behavior transition probability data, 28B Elapsed time probability distribution data, 28C Start time probability distribution data, 28D End time probability distribution data, 28E Behavior duration probability distribution data 30 Display section 32 Communication I / F 39 Bus 40 Feature extraction unit 42 Behavior estimation section 44 Information extraction part 46 Behavior Prediction Department 48 Behavior occurrence and behavior start time prediction section 50 Action end time prediction unit 60 Past behavioral sequence 62 Current Activities

Claims

1. A behavior prediction device that predicts a future behavior of a specific person, an information extraction unit that extracts, based on a past behavior sequence that represents the specific person's past behavior history in a time series, start time probability distribution data that represents a probability distribution of start times for each type of behavior, behavior transition probability data that represents the probability of transitioning from one behavior to another for each combination of behavior types, and behavior time data related to the behavior time for each type of behavior; a behavior prediction unit that predicts a future behavior of the specific person based on the start time probability distribution data, the behavior transition probability data, the behavior time data, and the current behavior of the specific person; A behavior prediction device comprising:

2. The behavior time data includes behavior time probability distribution data that represents a probability distribution of behavior time for performing a behavior for each type of behavior, and elapsed time probability distribution data that represents a probability distribution of elapsed time from the start of one behavior to the start of the other behavior for each combination of behavior types for each type of next behavior. The behavior prediction device according to claim 1 .

3. The system further includes a behavior estimation unit that estimates the current behavior of the specific person based on a feature obtained from a detection result of a sensor that detects an event that changes due to the person's behavior. The behavior prediction device according to claim 1 .

4. the information extraction unit further extracts, based on the past behavioral sequence, end time probability distribution data representing a probability distribution of end times for each type of behavior; The behavior prediction unit predicts future behavior of the specific person further based on the end time probability distribution data. The behavior prediction device according to claim 1 .

5. A behavior prediction method for predicting future behavior of a specific person using a behavior prediction device, comprising: an information extraction unit of the behavior prediction device extracts, based on a past behavior sequence that represents a time series of the specific person's past behavior history, start time probability distribution data that represents a probability distribution of start times for each type of behavior, behavior transition probability data that represents a probability of transitioning from one behavior to another for each combination of behavior types, and behavior time data related to behavior times for each type of behavior; a behavior prediction unit of the behavior prediction device predicts a future behavior of the specific person based on the start time probability distribution data, the behavior transition probability data, the behavior time data, and the current behavior of the specific person; Behavioral prediction methods.

6. A behavior prediction program for causing a computer to function as each unit of the behavior prediction device according to any one of claims 1 to 4.

Images