JP5590301B2 - Behavior determination device, behavior determination system, terminal device, behavior determination method, and program - Google Patents

Description

  The present invention relates to a behavior determination device, a behavior determination system, a terminal device, a behavior determination method, and a program used for determining a behavior of a person or the like.

  In recent years, using a sensor carried or worn by a person, the person's behavior is sensed, and from the obtained sensor data, the behavior of the person to be measured (target person) is determined, thereby managing health. Or, a service for recommending information has been attempted. In addition, such service attempts are increasing year by year.

  An acceleration sensor is mentioned as a sensor utilized in order to perform the above-mentioned service. When determining the behavior of the subject using the acceleration sensor, the acceleration sensor is carried by the subject or attached to the subject's body at a fixed position. In the action determination, first, data for a predetermined time is cut out from the obtained acceleration sensor data, and a feature amount is calculated from the cut out data. Next, the calculated feature amount is compared with a preset threshold value, and the actual action of the subject is determined from the relationship between the two. In general, it is determined which of a plurality of actions to be determined such as “walking”, “running”, and “stairs up / down” corresponds to. Hereinafter, actions involving movement, such as “walking”, “running”, and “step up / down” are collectively referred to as “actions”.

  Patent document 1 is mentioned as literature which discloses the example of the system which determines such a general action. Patent Document 1 discloses a system for determining an action such as “walking” or “step up” of an object person using an acceleration sensor fixed to the waist part of the object person. The system disclosed in Patent Document 1 determines the behavior of the subject by obtaining the vibration intensity of the vibration generated when the subject acts and the movement of the center of gravity from the acceleration sensor data.

  However, in the system disclosed in Patent Document 1, the subject needs to mount the sensor at a designated position, or explicitly declare the mounting position to the system administrator. As described above, it is inconvenient and burdensome for the subject to mount the sensor at a predetermined position or to declare the mounting position.

  On the other hand, in the system disclosed in Patent Document 1, if the subject's sensor mounting position is unknown, it is necessary to determine the behavior of the subject only from the vibration intensity and the movement of the center of gravity. The accuracy of the remarkably decreases, and the determination becomes difficult. This is because the magnitude of the value of the acceleration sensor data varies depending on the position where the subject wears the sensor even if the behavior is the same. Therefore, when using the system disclosed in Patent Document 1, it is necessary to clarify the sensor mounting position of the subject from the viewpoint of determination accuracy.

  In addition, Patent Document 2 discloses an example of a system that determines the movement of a subject in the vertical direction (vertical direction) using peak information of acceleration sensor data. “Peak information” is information relating to the peak of the mountain shape of the acceleration sensor data. In the present specification, the peak of the peak shape of the sensor data is hereinafter referred to as “peak”.

  In general, it is known that a peak appears in acceleration sensor data worn on the body every time a person steps one foot. Although the magnitude of the peak value varies depending on the mounting position of the acceleration sensor, it is a fact that the occurrence of the peak represents the grounding of the foot regardless of the mounting position. Therefore, by using the presence / absence of the occurrence of the peak and the occurrence interval thereof, the presence / absence of the grounding of the foot and the period of the grounding, i.e., the presence / absence of the movement of the subject, and the pace thereof are determined regardless of the sensor mounting position. Can get to.

  Based on such a principle, the system disclosed in Patent Document 2 obtains the peak of acceleration sensor data worn by the subject, and integrates the vertical acceleration between the obtained peak. Then, the system disclosed in Patent Document 2 determines whether the obtained integrated value takes a positive or negative value, thereby determining the movement of the subject in the vertical direction, that is, the stair climbing action or the like. .

  However, the system disclosed in Patent Document 2 has a problem that the determination accuracy is low. This is because the acceleration data generated along with the action of the target person includes noise components generated due to various factors, and the influence thereof cannot be ignored. That is, the noise component generated when the target person acts causes a random bias in the vertical direction in the acceleration sensor data. For this reason, it is difficult to distinguish ascending, descending, and flat ground movement only from the positive / negative integral value in the vertical direction of the acceleration sensor data or the magnitude relationship thereof, and the determination accuracy is significantly reduced.

  Further, the system disclosed in Patent Document 2 detects a change point of a moving action in the vertical direction, and determines that the action of the target person is assumed after the change point that the action is continued. This is because, in the system disclosed in Patent Document 2, the integrated value of the vertical acceleration during movement is considered to be substantially equal to 0 (zero).

However, such a premise is equivalent to the assumption that the movement in the vertical direction is a uniform motion, and does not necessarily match the actual movement behavior. Furthermore, under this assumption, when an erroneous detection of a change point occurs, all subsequent behavior determinations are erroneous determinations. And in the long-time measurement, since the influence of such misjudgment accumulates, there is a possibility that the judgment accuracy is remarkably lowered.
(For example, refer to Patent Document 1.)

Japanese Patent Application Laid-Open No. 2007-074428 JP 2006-170879 A

  As described above, the system disclosed in Patent Document 1 has a problem in that it imposes inconvenience and burden on the subject because there are restrictions on how to attach the sensor, and further attempts to solve this problem. Then, the determination accuracy is lowered. Further, the system disclosed in Patent Document 2 has a problem that the determination accuracy is low because the influence of the noise component cannot be removed particularly in the determination of the movement of the subject in the vertical direction.

  An object of the present invention is to solve the above-mentioned problems, is not subject to restrictions on how to attach the sensor, and can perform a behavior determination with high accuracy even when the determination target moves in the vertical direction. To provide a determination system, a terminal device, a behavior determination method, and a program.

In order to achieve the above object, an action determination apparatus according to the present invention is an action determination apparatus that determines an action to be measured based on time-series sensor data output from an acceleration sensor,
A time window extraction unit that extracts time window data of a set time length from the sensor data; and
A feature amount calculation unit that calculates a feature amount that serves as an index of behavior evaluation of the measurement target;
An action determination unit for determining an action of the measurement target based on the feature amount;
With
The feature amount calculation unit extracts peak information for specifying a peak included in the time window data from the time window data, specifies one peak and a peak immediately before the peak information based on the peak information, A portion existing between two specified peaks in the sensor data is cut out as peak-to-peak data, and from the peak-to-peak data, the vertical acceleration applied to the measurement object is equal to or less than a set value as the feature amount. The degree of floating indicating the degree of occurrence of the state is calculated.

In order to achieve the above object, an action determination system according to the present invention determines an action of a measurement object based on an acceleration sensor that outputs time-series sensor data according to the action of the measurement object, and the sensor data. An action determination device,
The behavior determination device is configured to extract, from the sensor data, a time window extraction unit that extracts time window data of a set time length, and a feature amount calculation unit that serves as an index for behavior evaluation of the measurement target. And an action determination unit that determines an action of the measurement target based on the feature amount,
The feature amount calculation unit extracts peak information for specifying a peak included in the time window data from the time window data, specifies one peak and a peak immediately before the peak information based on the peak information, A portion existing between two specified peaks in the sensor data is cut out as peak-to-peak data, and from the peak-to-peak data, the vertical acceleration applied to the measurement object is equal to or less than a set value as the feature amount. The degree of floating indicating the degree of occurrence of the state is calculated.

Furthermore, in order to achieve the above object, the terminal device according to the present invention provides:
An acceleration sensor that outputs time-series sensor data according to the behavior of the measurement target;
A time window extraction unit that extracts time window data of a set time length from the sensor data; and
A feature amount calculation unit that calculates a feature amount that serves as an index of behavior evaluation of the measurement target;
An action determination unit for determining an action of the measurement target based on the feature amount;
With
The feature amount calculation unit extracts peak information for specifying a peak included in the time window data from the time window data, specifies one peak and a peak immediately before the peak information based on the peak information, A portion existing between two specified peaks in the sensor data is cut out as peak-to-peak data, and from the peak-to-peak data, the vertical acceleration applied to the measurement object is equal to or less than a set value as the feature amount. The degree of floating indicating the degree of occurrence of the state is calculated.

In order to achieve the above object, an action determination method in the present invention is a method for determining an action to be measured based on time-series sensor data output from an acceleration sensor,
(A) cutting out time window data of a set time length from the sensor data; and
(B) extracting from the time window data peak information identifying a peak included in the time window data;
(C) identifying one peak and the immediately preceding peak based on the peak information, and cutting out a portion existing between the two specified peaks in the sensor data as inter-peak data;
(D) From the peak-to-peak data, as a feature quantity serving as an index for evaluating the behavior of the measurement target, a degree of float indicating a degree of occurrence of a state in which a vertical acceleration applied to the measurement target is equal to or less than a set value is calculated. Steps,
(E) determining the behavior of the measurement object based on the feature amount; and
It is characterized by having.

In order to achieve the above object, a program according to the present invention is a program for executing a determination of an action of a measurement object based on time-series sensor data output from an acceleration sensor by a computer,
In the computer,
(A) cutting out time window data of a set time length from the sensor data; and
(B) extracting from the time window data peak information identifying a peak included in the time window data;
(C) identifying one peak and the immediately preceding peak based on the peak information, and cutting out a portion existing between the two specified peaks in the sensor data as inter-peak data;
(D) From the peak-to-peak data, as a feature quantity serving as an index for evaluating the behavior of the measurement target, a degree of float indicating a degree of occurrence of a state in which a vertical acceleration applied to the measurement target is equal to or less than a set value is calculated. Steps,
(E) determining the behavior of the measurement object based on the feature amount; and
Is executed.

  As described above, according to the present invention, it is possible to perform behavior determination with high accuracy even when the determination target is not restricted and the determination target moves in the vertical direction.

FIG. 1 is a diagram schematically showing the vertical movement that occurs in the body when a person moves. FIG. 1A shows a case where a person is walking on a flat ground, while FIG. FIG. 1C shows the time when the person is going down the stairs. FIG. 2 is a diagram for explaining the “floating degree” in the present invention, and shows an example of sensor data. FIG. 3 is a block diagram illustrating a configuration of the behavior determination apparatus and the behavior determination system according to Embodiment 1 of the present invention. FIG. 4 is a diagram illustrating an example of a determination rule when an acceleration dispersion value is used as a feature amount. FIG. 5 is a diagram illustrating an example of a determination rule when the peak interval is used as the feature amount. FIG. 6 is a diagram illustrating an example of a determination rule when the degree of floating (depth area between peaks) is used as the feature amount. FIG. 7 is a diagram illustrating another example of the determination rule when the acceleration dispersion value is used as the feature amount. FIG. 8 is a diagram illustrating another example of the determination rule when the peak interval is used as the feature amount. FIG. 9 is a diagram illustrating an example of a determination rule when the degree of floating (depth area between peaks) is used as the feature amount. FIG. 10 is a block diagram illustrating another example of the terminal device and the behavior determination device according to Embodiment 1 of the present invention. FIG. 11 is a flowchart showing the operation of the behavior determination apparatus according to Embodiment 1 of the present invention. FIG. 12 is a block diagram illustrating a configuration of the behavior determination device and the behavior determination system according to Embodiment 2 of the present invention. FIG. 13 is a diagram illustrating an example of a determination rule used in Embodiment 2 of the present invention. FIG. 14 is a flowchart showing the operation of the behavior determination apparatus according to Embodiment 2 of the present invention. FIG. 15 is a block diagram illustrating a configuration of the behavior determination device and the behavior determination system according to Embodiment 3 of the present invention. FIG. 16 is a flowchart showing the operation of the behavior determination apparatus according to Embodiment 3 of the present invention. FIG. 17 is a block diagram showing a configuration of the behavior determination apparatus according to Embodiment 4 of the present invention. FIG. 18 is a diagram illustrating an example of a screen of the terminal device when the user is requested to input whether the determination result is correct in Embodiment 4 of the present invention. FIG. 19 is a diagram illustrating an example of a determination rule changed after learning in the fourth embodiment of the present invention. FIG. 20 is a flowchart showing the operation of the behavior determination apparatus according to Embodiment 4 of the present invention. FIG. 21 is a block diagram illustrating the configuration of the behavior determination device and the behavior determination system according to the fifth embodiment of the present invention. FIG. 22 is a flowchart showing the operation of the behavior determination apparatus according to Embodiment 5 of the present invention. FIG. 23 is a block diagram illustrating an example of a computer that can execute the program according to the first to fifth embodiments of the present invention. FIG. 24 is a diagram illustrating an example of sensor data output by the acceleration sensor in the first embodiment. FIG. 25 is a diagram illustrating an example of sensor data output from the acceleration sensor in the second embodiment. FIG. 26 is a diagram illustrating an example of a determination rule used in the second embodiment.

(Outline of the present invention)
The present invention uses a feature that appears in sensor data that is output from an acceleration sensor from the time when one foot of a measurement object, for example, a person touches down, until the other foot touches the ground in the next step. And this invention catches the state where vertical acceleration (typically gravitational acceleration) that is steadily acting on the measurement object is reduced from the features that appear in the sensor data, and the degree of occurrence of this state. Use to determine the action. For this reason, in this invention, even if it is a case where a measuring object moves to an up-down direction, without receiving restrictions regarding how to attach an acceleration sensor, action can be determined with sufficient precision.

  Here, the principle of the present invention will be described with reference to FIGS. FIG. 1 is a diagram schematically showing the vertical movement that occurs in the body when a person moves. FIG. 1A shows a case where a person is walking on a flat ground, while FIG. FIG. 1C shows the time when the person is going down the stairs. FIG. 2 is a diagram for explaining the “floating degree” in the present invention, and shows an example of sensor data.

  As shown in FIGS. 1A to 1C, when a person moves, a vertical movement of the body accompanying the raising and lowering of the foot occurs. When moving up and down, such as when going up and down stairs (see FIGS. 1B and 1C), compared to walking on a flat ground (FIG. 1A), there is an action of raising or lowering the foot. Occur.

  The method of raising and lowering these feet is different between the movements when climbing the stairs, the movements when descending the stairs, and the movement on the flat ground. The difference between these movements is that the next step is taken after one foot is stepped on. Thus, it can be characterized by the degree of occurrence of a decrease in acceleration in the vertical direction applied to the body until the other foot reaches the ground. Generally, there is a tendency that the acceleration in the vertical direction decreases in the order of movement on flat ground, movement due to ascending stairs, and movement due to descending stairs. When the acceleration in the vertical direction is decreasing, the person feels a floating feeling at the center of gravity. In other words, the person feels like floating in the air.

  By the way, acceleration is constantly applied to the acceleration sensor in the vertical direction. A representative example of acceleration that constantly acts in the vertical direction is gravitational acceleration. Here, when only the gravitational acceleration is considered as the acceleration acting constantly, the magnitude of the acceleration is “1G”. Therefore, when the acceleration sensor outputs sensor data indicating a value of 1 G or less, the person who is the measurement target of this acceleration sensor feels a floating feeling. That is, at this time, the acceleration sensor can be regarded as a state floating in the air.

  Therefore, the degree of occurrence of the state in which the acceleration in the vertical direction is reduced is, for example, the time during which sensor data of 1G or less is output between the time when one foot is grounded and the other foot is grounded. Or how much smaller than 1G. Then, the point of time when the foot is touched by one step is almost synchronized with the point of time when the peak of the sensor data from the acceleration sensor occurs.

  For this reason, for example, when only gravity is considered as a steady acceleration, the degree of occurrence of a state in which the vertical acceleration applied to the measurement target is reduced can be represented by the feature amount (shaded portion) shown in FIG. . Specifically, for example, as shown in FIG. 2, the area of the region of 1 G or less between peaks (the area of the hatched portion in FIG. 2) is the state in which the vertical acceleration applied to the measurement object is reduced. The degree of occurrence is shown. Therefore, it can be determined from the degree of the hatched area in FIG. 2 whether the measurement object is going up the stairs, going down the stairs, or moving on the flat ground.

(Embodiment 1)
Hereinafter, a behavior determination device, a behavior determination system, a terminal device, a behavior determination method, and a program according to Embodiment 1 of the present invention will be described with reference to FIGS. Initially, the structure of the action determination apparatus and action determination system in this Embodiment 1 is demonstrated using FIG. FIG. 3 is a block diagram illustrating a configuration of the behavior determination apparatus and the behavior determination system according to Embodiment 1 of the present invention.

  The behavior determination device 10 illustrated in FIG. 1 is a device that determines a measurement target behavior based on time-series sensor data output from the acceleration sensor 2. The behavior determination system 100 includes a behavior determination device 10. As illustrated in FIG. 1, the behavior determination device 10 includes a time window extraction unit 11, a feature amount calculation unit 12, and a behavior determination unit 16.

  The time window extraction unit 11 extracts time window data having a set time length from the sensor data. The feature amount calculation unit 12 calculates a feature amount that serves as an index for behavior evaluation of a measurement target. The feature amount calculation unit 12 includes a peak extraction unit 13, a peak-to-peak data cutout unit 14, and a floating degree calculation unit 15 in order to calculate a feature amount.

  In the feature amount calculation unit 12, the peak extraction unit 13 extracts peak information that identifies a peak included in the time window data from the time window data extracted by the time window extraction unit 11. The peak-to-peak data extraction unit 14 identifies one peak and the peak immediately preceding it based on the peak information extracted by the peak extraction unit 13, and is a portion existing between the two identified peaks in the sensor data Are extracted as peak-to-peak data.

  The floating degree calculation unit 15 calculates, from the peak-to-peak data cut out by the peak-to-peak data cutout unit 14, the degree of occurrence of a state in which the vertical acceleration applied to the measurement target is equal to or less than a set value as a feature amount. The set value of acceleration is set to, for example, the value of gravitational acceleration (1G). However, the set value of the acceleration is not limited to 1G, and is set as appropriate according to the condition of the measurement target.

  The behavior determination unit 16 determines the behavior of the measurement target based on the feature amount, in particular, “the degree of occurrence of a state where the vertical acceleration applied to the measurement target is equal to or less than the set value”. In the first embodiment, as will be described later, the behavior determination unit 16 uses other feature quantities in addition to “the degree of occurrence of a state in which the vertical acceleration applied to the measurement target is equal to or less than the set value”. Thus, the behavior to be measured can also be determined.

  In this way, the behavior determination apparatus 10 can determine the behavior of a measurement target, for example, a person, using “the degree of occurrence of a state in which the vertical acceleration applied to the measurement target is equal to or less than a set value” as a feature amount. it can. As described above, the “degree of occurrence of a state in which the vertical acceleration applied to the measurement object is equal to or less than the set value” corresponds to the degree of occurrence of a state in which the measurement object feels floating while moving.

  Therefore, according to the behavior determination apparatus 10, it is possible to determine the behavior with high accuracy even when the measurement object moves in the vertical direction without being restricted with respect to how to attach the acceleration sensor. Further, hereinafter, “the degree of occurrence of a state in which the vertical acceleration applied to the measurement object is equal to or less than the set value” is referred to as “floating degree”.

  Here, the configurations of the behavior determination device 10 and the behavior determination system 100 will be described more specifically. As shown in FIG. 1, in the first embodiment, the behavior determination system 100 includes a behavior determination device 10 and a terminal device 1. The behavior determination apparatus 10 is constructed by a server computer. A specific example of the terminal device 1 is an electronic device carried by a person such as a mobile phone. The terminal device 1 includes an acceleration sensor 2, a data acquisition unit 3, and an output device 4. Although not shown, the terminal device 1 is also provided with an input device such as a keyboard, a touch panel, or both.

  The data acquisition unit 3 has a function of acquiring time-series sensor data output from the acceleration sensor 2 and holding a certain amount of data. The data acquisition unit 3 also has a communication function for transmitting sensor data acquired from the acceleration sensor 2 to the behavior determination device 10. The output device 4 is a device that indicates a result of behavior determination to the user, such as a display device.

  As described above, in the behavior determination system 100 according to the first embodiment, the acceleration sensor 2 is mounted on the terminal device 1 (mobile phone) held by the person to be measured, and the sensor data output by the acceleration sensor 2 is the terminal It is transmitted to the action determination device 10 using the communication function of the device 1. Then, the server computer functioning as the behavior determination device 10 performs behavior determination, and the result is returned to the terminal device 1 possessed by the person and notified to the user by the output device 4.

  Moreover, in this Embodiment 1, the action determination apparatus 1 is provided with the output part 17 in addition to the time window extraction part 11, the feature-value calculation part 12, and the action determination part 16. FIG. The output unit 17 outputs the result of determination by the behavior determination unit 16 to the outside (terminal device 1). In the eleventh embodiment, the output unit 17 has a function of communicating with the terminal device 1.

  In the behavior determination device 10, the time window extraction unit 11 extracts time window data having a set time length from the sensor data output from the data acquisition unit 3 as described above, and extracts the extracted time window data. Is output to the peak extraction unit 13 that performs peak extraction.

  In the first embodiment, the time length (time window length) of the time window data extracted by the time window extraction unit 11 is determined in advance, such as 1 second. In the first embodiment, the method of extracting time window data by the time window extracting unit 11 is not particularly limited. Here, when the time window cutout unit 11 cuts out time window data at a specific time, for example, consider a case where the time window is set to a length of 1 second and cut out. In this case, the time window extraction unit 11 may extract data of 0.5 seconds before and after the center of a certain time, and may extract time window data having a length of 1 second. As the end, data for the past 1 second may be cut out, and a time window of 1 second in length may be cut out.

  Further, when the data acquisition unit 3 continuously receives sensor data from the acceleration sensor 2, the time window extraction unit 11 divides the continuous sensor data by a time window of a predetermined length. By doing so, the time window data can also be cut out. Specifically, the time window cutting unit 11 sets the first time window from time 0 second to time 1 second, sets the second time window from time 1 second to time 2 seconds, and sets each time window set. Extract time window data in the time window. Furthermore, in this case, the time window extraction unit 11 can also extract time window data so that adjacent time windows have overlapping portions of a certain interval. For example, the time window cutout unit 11 can set the first time window from time 0 seconds to time 1 second, and can set the second time window from time 0.5 seconds to time 1.5 seconds. .

  As described above, a plurality of different modes can be considered for the method of extracting the time window data by the time window extracting unit 11. In the first embodiment, an appropriate clipping method is adopted according to the state, length, etc. of sensor data. Also, in other embodiments described later, as in the above-described example, an appropriate one of a plurality of cutting methods is employed.

  As described above, the peak extraction unit 13 uses the time window data provided from the time window extraction unit 11 to extract peak information that identifies the peak included in the time window data. In the first embodiment, for example, the peak extraction unit 13 can regard a data point indicating the maximum value in one time window data as a peak. A plurality of peak information may be extracted from one time window data, and when no peak exists, the peak information itself may not be output. The peak information includes the sensor data value (output value) at the peak and the time.

  Moreover, the peak extraction part 13 can also attach the following conditions at the time of peak specification, and can also improve the precision of peak extraction. For example, the peak extraction unit 13 creates a polygonal line by connecting the data point of the maximum value in the time window data and the data points before and after the data line with a straight line, and the shape of the polygonal line has the data point of the maximum value as a vertex. If the shape is convex upward, the maximum data point can be regarded as a peak. Further, the peak extraction unit 13 determines the maximum value data point when the maximum acceleration in the time window data is a certain value or more, or when the acceleration dispersion value calculated from the time window data is a certain value or more. Can also be considered a peak.

  In addition, the peak extraction unit 13 specifies the second largest data point in addition to the maximum data point in the time window data, and applies the above condition to the two obtained data points. Either one can be regarded as a peak. Furthermore, the peak extraction unit 13 can also regard the data point having the second largest value as a peak on condition that the time interval between the two obtained data points is a predetermined interval or more.

  Further, in the first embodiment, the method of peak extraction in the peak extraction unit 13 is not limited to the above-described mode, and the acceleration sensor data is obtained by touching the foot when the person to be measured takes a step. Any method can be used as long as the peak that appears can be acquired with high accuracy. Further, in other embodiments described later, as in the above-described example, an appropriate one is employed from among a plurality of extraction methods.

  Furthermore, in the first embodiment, the peak extraction unit 13 calculates a feature amount other than the above-described float degree. Examples of the feature amount other than the floating degree include statistics such as an average value and a variance value obtained from each time window data, kurtosis, skewness, FFT power spectrum, an interval between two specified peaks, and the like. When the peak extraction unit 13 calculates a feature amount other than the floating degree, it outputs this to the behavior determination unit 16.

  As described above, the peak-to-peak data cutout unit 14 identifies the latest peak and the peak immediately preceding it from the time of occurrence of each peak based on the peak information obtained from the peak extraction unit 13. Then, the peak-to-peak data cutout unit 14 cuts out the identified peak-to-peak sensor data from the sensor data acquired from the data acquisition unit 3. Further, the peak-to-peak data cutout unit 14 outputs the cut-out peak-to-peak data to the floating degree calculation unit 15.

  In the first embodiment, the method of extracting the peak-to-peak data in the peak-to-peak data extraction unit 14 is not particularly limited. Here, consider a case where the data acquisition unit 3 has a function of storing sensor data for a certain period of time, for example. In this case, the peak-to-peak data extraction unit 14 can also acquire sensor data in a time zone corresponding to the peak from the data acquisition unit 3 based on the peak information.

  Further, the peak-to-peak data cutout unit 14 itself may have a function of storing sensor data. In this case, the sensor data for a certain time given from the data acquisition unit 3 is accumulated. When the latest peak information is acquired from the peak extraction unit 13, the peak-to-peak data extraction unit 14 can extract sensor data in a time zone corresponding to the peak from the stored sensor data. Further, in other embodiments described later, as in the above-described example, an appropriate one of a plurality of cutting methods can be employed.

  As described above, the floating degree calculation unit 15 uses the peak-to-peak data provided from the peak-to-peak data cutout unit 14, and the degree of occurrence of the state in which the vertical acceleration applied to the measurement target is equal to or less than the set value (floating Degree). Further, the floating degree calculation unit 15 outputs the calculated floating degree to the behavior determination unit 16.

  Here, for example, it is assumed that only gravitational force is considered as an acceleration component that acts constantly, and the acceleration setting value is set to the value of gravitational acceleration (1G). In this case, as the degree of floating, the area (shaded portion) of a region of 1 G or less between the peaks shown in FIG. The buoyancy calculation unit 15 integrates the portion of the peak-to-peak data that is less than or equal to the set value (1G or less) over time, and the area of the region that is less than or equal to the set value between the peaks (hereinafter referred to as “depth between peaks”). Is calculated. The set value is not limited to 1G, and is set as appropriate according to the environment where the measurement target exists. A specific calculation process of the peak-to-peak depth area will be described later.

  In the first embodiment, the degree of floating is not limited to the peak-to-peak depth area. Examples other than the peak-to-peak depth area include the section length of a region below a set value between peaks, or the average value of acceleration sensor data in that region. Further, when an acceleration component other than gravity is assumed as an acceleration component that acts constantly, the above-described “set value” is set in consideration of the assumed acceleration component.

  The behavior determination unit 16 can determine the behavior to be measured based on the feature amount calculated by the feature amount calculation unit 12 as described above, specifically, based on the degree of floating. In the first embodiment, the behavior determination unit 16 determines which of the plurality of behaviors that the behavior to be measured corresponds to in advance according to the determination rule. As a determination rule, a threshold value is set for each assumed behavior in the value of the peak-to-peak area depth used as the floating degree, and the behavior determining unit 15 determines the behavior based on the threshold value (see FIG. 6).

  Further, as described above, the degree of floating corresponds to the degree of occurrence of a state in which the measurement object feels floating while moving. Therefore, the behavior assumed in the determination based on the degree of floating includes behavior with a large movement in the vertical direction, such as climbing stairs, descending stairs, skipping, jumping, climbing hills, descending hills, etc. .

  In the first embodiment, for example, the behavior determination unit 16 can also perform behavior determination using “floating degree” and other feature amounts as feature amounts. Here, with reference to FIGS. 4 to 6, the action determination unit 16 performs “the degree of float”, “the acceleration dispersion value”, and the “interval between two peaks extracted by the peak extraction unit 13 (hereinafter referred to as“ peak interval ”). ) ”Will be described. In the following example, there are five actions that are assumed in advance: “stop”, “run”, “walk”, “up stairs”, and “down stairs”.

  FIG. 4 is a diagram illustrating an example of a determination rule when an acceleration dispersion value is used as a feature amount. When the behavior determination unit 16 acquires the acceleration variance value as the feature amount, the behavior determination unit 16 applies this value to the determination rule illustrated in FIG. 4 and determines whether the behavior to be measured corresponds to “stop”. In the example of FIG. 4, when the variance value is less than 4000 [mG · mG], the behavior determination unit 16 determines that the behavior to be measured corresponds to “stop”. On the other hand, when the variance value is 4000 [mG · mG] or more, the behavior determination unit 16 determines that the behavior to be measured corresponds to “movement (including all behaviors other than stopping)”.

  FIG. 5 is a diagram illustrating an example of a determination rule when the peak interval is used as the feature amount. When the behavior determination unit 16 acquires the peak interval as the feature amount, the behavior determination unit 16 applies this to the determination rule illustrated in FIG. 5 and determines whether the behavior to be measured corresponds to “run” or “walk”. The peak interval represents the pace of the person to be measured. When the peak interval is shortened, the pace is shortened and the possibility that the person is running increases. Therefore, in the example of FIG. 5, when the peak interval is less than 500 [msec], the behavior determination unit 16 determines that the behavior to be measured corresponds to “run”. On the other hand, when the peak interval is 500 [msec] or more, the behavior determining unit 16 determines that the behavior to be measured corresponds to “walking”.

  FIG. 6 is a diagram illustrating an example of a determination rule when the degree of floating (depth area between peaks) is used as the feature amount. In the example of FIG. 6, the peak-to-peak depth area shown in FIG. 2 is used as the floating degree. When the behavior determination unit 16 obtains the peak-to-peak depth area as the degree of float that is a feature amount, the behavior determination unit 16 applies this to the determination rule illustrated in FIG. ”Or“ step down ”is determined.

  In the example of FIG. 6, when the depth area between peaks is less than 70 [mG · sec], the behavior determination unit 16 determines that the behavior to be measured corresponds to “walking (moving on a flat ground)”. Moreover, when the depth area between peaks is 70 [mG · sec] or more and less than 90 [mG · sec], the action determination unit 16 determines that the action to be measured corresponds to “step up”. Further, when the depth area between the peaks is 90 [mG · sec] or more and less than 150 [mG · sec], the action determination unit 16 determines that the action to be measured corresponds to “step down”. Specifically, if the given peak-to-peak depth area is, for example, 85 [mG · sec], the action determination unit 16 determines that the action to be measured is “step up” from the determination rule shown in FIG. Is determined.

  Moreover, in this Embodiment 1, the action determination part 16 can also perform determination using all the determination rules shown in FIGS. 4-6, in order to identify the measurement object action. Specifically, the behavior determination unit 16 first applies the acceleration variance value to the determination rule shown in FIG. Next, when it is determined that the measurement target is not stopped and “moving”, the behavior determination unit 16 continues to apply the peak interval to the determination rule illustrated in FIG. 5. Next, when the action determination unit 16 determines that the action to be measured is “walking”, the action determination unit 16 continues to apply the peak-to-peak depth area to the determination rule illustrated in FIG. 6.

  When the action determination unit 16 performs the above-described processing using all the determination rules shown in FIGS. 4 to 6, only one action is selected from a plurality of action candidates assumed in advance. The action determination unit 16 determines that the action to be measured is the selected action. Thereafter, the behavior determination unit 16 outputs the obtained determination result to the output unit 17.

  Moreover, in this Embodiment 1, the action determination part 16 can also perform action determination using determination rules other than the determination rule shown in FIGS. Below, the other example of the action determination by the action determination part 16 is demonstrated using FIGS.

  When the determination rules shown in FIGS. 7 to 9 are used, the behavior determination unit 16 calculates an evaluation value for each assumed behavior. The evaluation value indicates the possibility that the behavior to be measured corresponds to the behavior corresponding to each rule. In the following example, there are five presumed actions: “stop”, “run”, “walk”, “up stairs”, and “down stairs”.

  FIG. 7 is a diagram illustrating another example of the determination rule when the acceleration dispersion value is used as the feature amount. When the behavior determination unit 16 acquires the acceleration dispersion value as the feature amount, the behavior determination unit 16 applies this value to the determination rule illustrated in FIG. 7 and calculates an evaluation value indicating the possibility that the behavior to be measured corresponds to “stop”. The rule shown in FIG. 7 defines the relationship between the acceleration dispersion value and the degree corresponding to “stop”. In FIG. 7, the evaluation value on the vertical axis represents the degree to which the action to be measured corresponds to “run” by a numerical value from 0 to 100.

From FIG. 7, it is assumed that the evaluation value representing the possibility of being “stopped” is Ds and the acceleration dispersion value is V. When the acceleration dispersion value V is 0 [mmG ² ] or more and less than 4000 [mmG ² ], the evaluation value is obtained. Ds is expressed by the following (Equation 1). When V is 4000 [mmG ² ] or more, the evaluation value Ds is expressed by the following (Equation 2).

(Equation 1)
Ds = − (1/40) × V + 100

(Equation 2)
Ds = 0

  FIG. 8 is a diagram illustrating another example of the determination rule when the peak interval is used as the feature amount. When the behavior determination unit 16 acquires the peak interval as the feature amount, the behavior determination unit 16 applies this to the determination rule illustrated in FIG. 8 and calculates an evaluation value indicating the possibility that the behavior to be measured corresponds to “run” or “walk”. . The rule shown in FIG. 8 defines the relationship between the peak interval and the degree corresponding to “run” or “walk”. In FIG. 8, the evaluation value on the vertical axis represents the degree to which the action to be measured corresponds to “running” or “walking” as a numerical value from 0 to 100.

  From FIG. 8, when the evaluation value indicating the possibility of being “walking” is Dw and the peak interval is P, when the peak interval P is 600 [msec] or more and less than 3000 [msec], the evaluation value Dw is as follows: (Expression 3) When P is less than 600 [msec], the evaluation value Dw is expressed by the following (Equation 4). Further, when the peak interval is 3000 [msec] or more, the evaluation value Dw is expressed by the following (Equation 5).

(Equation 3)
Dw = (− 1/24) × P + 125

(Equation 4)
Dw = 100

(Equation 5)
Dw = 0

  Further, from FIG. 8, when an evaluation value indicating the possibility of being “running” is set to Dr, the evaluation value is obtained when the peak interval P is not less than 300 [msec] and less than 500 [msec], similarly to “walking”. Dr is expressed by the following (Equation 6). When the peak interval P is less than 300 [msec], the evaluation value Dr is expressed by the following (Equation 7). Furthermore, when the peak interval P is 500 [msec] or more, the evaluation value Dr is expressed by the following (Equation 8).

(Equation 6)
Dr = − (1/2) × P + 250

(Equation 7)
Dr = 100

(Equation 8)
Dr = 0

  FIG. 9 is a diagram illustrating an example of a determination rule when the degree of floating (depth area between peaks) is used as the feature amount. In the example of FIG. 9 as well, as in the example of FIG. 6, the peak-to-peak depth area shown in FIG. When the behavior determination unit 16 obtains the peak-to-peak depth area as the floating degree that is the feature amount, this is applied to the determination rule illustrated in FIG. 9, and the behavior to be measured is “step up” or “step down”. It is determined whether it falls under any of the above, or further.

  The rule shown in FIG. 9 defines the relationship between the peak-to-peak depth area and the degree corresponding to “step up” or “step down”. In FIG. 9, the evaluation value on the vertical axis represents the degree to which the action to be measured corresponds to “ascending stairs” or “descending stairs” as a numerical value from 0 to 100. Further, according to the rule shown in FIG. 9, when the peak-to-peak depth area is less than 70 [mG · sec], the behavior determination unit 16 determines whether the behavior to be measured is “up stairs” or “down stairs”. It is determined that it corresponds to “walking (moving on flat ground)”.

  From FIG. 9, when the evaluation value indicating the possibility of being “up the stairs” is Du and the floating degree is F, when the floating degree F is 70 [mG · sec] or more and less than 80 [mG · sec], The evaluation value Du is expressed by the following (Equation 9). When the floating degree F is 80 [mG · sec] or more and less than 90 [mG · sec], the evaluation value Du is expressed by the following (Equation 10). Furthermore, when the floating degree F is less than 70 [mG · sec] or 90 [mG · sec] or more, the evaluation value Du is expressed by the following (Equation 11).

(Equation 9)
Du = 10 × F-700

(Equation 10)
Du = −10 × F + 900

(Equation 11)
Du = 0

  Similarly, from FIG. 9, when an evaluation value indicating the possibility of falling down the stairs is Dd, when the floating degree F is 90 [mG · sec] or more and less than 120 [mG · sec], the evaluation is performed. The value Dd is expressed by the following (Equation 12). When the floating degree F is 120 [mG · sec] or more and less than 150 [mG · sec], the evaluation value Dd is represented by the following (Equation 13). Furthermore, when the floating degree F is less than 90 [mG · sec] or 150 [mG · sec] or more, the evaluation value Dd is expressed by the following (Equation 14).

(Equation 12)
Dd = (10/3) × F−300

(Equation 13)
Dd = − (10/3) × F + 500

(Equation 14)
Dd = 0

  In the first embodiment, when the determination rules shown in FIGS. 7 to 9 are used, the behavior determination unit 16 may output all the calculated evaluation values as a result of the behavior determination, and the evaluation value is 80, for example. Only actions showing the above values may be output as the result of action determination. In the former case, the degree of each action is displayed simultaneously. In this case, for example, if the degrees of “walking” and “running” are 50 and 50, respectively, based on the determination result, the user can determine that the action to be measured is between “walking” and “running”. Can be considered a state. Further, based on the degree of “running” and the degree of “ascending stairs”, the user can also determine that the measurement target is climbing up the stairs while running.

  Also, when the determination rules shown in FIGS. 7 to 9 are used, the action determination unit 16 selects only one action as the determination result, similarly to the case where the determination rules shown in FIGS. 4 to 6 are used. You can choose. Specifically, for example, the behavior determination unit 16 can identify the behavior showing the highest evaluation value among the evaluation values of each behavior, and can select this behavior as a result of the behavior determination.

  Note that the determination rule described above can be applied not only to the first embodiment but also to other embodiments described later. Also in other embodiment, the action determination part 16 can select and use the suitable determination rule according to the objective from the determination rules mentioned above.

  By the way, in the example mentioned above, although the action determination apparatus 10 is constructed | assembled by the apparatus different from the terminal device 1 which mounts the acceleration sensor 2, in this Embodiment 1, the action determination apparatus 10 is based on a terminal device. It may be built. This example will be described with reference to FIG. FIG. 10 is a block diagram illustrating another example of the terminal device and the behavior determination device according to Embodiment 1 of the present invention. In the example of FIG. 10, the behavior determination device 10 is built inside the terminal device 101. As illustrated in FIG. 10, in this example, the terminal device 101 includes all the configurations of the terminal device 1 and the behavior determination device 10 illustrated in FIG. 3.

  Although not shown in FIGS. 3 and 10, the first embodiment may be an embodiment in which the behavior determination device 10 shown in FIG. 3 is constructed by a computer, particularly a personal computer. In this aspect, the output device 4 may be a display device connected to the computer.

  And when it is set as such an aspect, the person of a measuring object will carry the apparatus provided with the acceleration sensor 2 and the data acquisition part 3. FIG. Furthermore, when it is set as such an aspect, as the data acquisition part 3, the apparatus which acquires sensor data in real time from an acceleration sensor, and transmits the acquired sensor data to the action determination apparatus 10 by radio | wireless communication is mentioned. The data acquisition unit 3 may be a device that stores sensor data from the acceleration sensor and transfers the stored sensor data to the behavior determination device 10 after the measurement is completed.

  The apparatus aspect described above can be applied not only to the first embodiment but also to other embodiments described later. Also in other embodiment, the action determination apparatus 10 may be constructed | assembled by the terminal device, and may be constructed | assembled by the computer.

  Next, operation | movement of the action determination apparatus 10 in Embodiment 1 of this invention is demonstrated using FIG. FIG. 11 is a flowchart showing the operation of the behavior determination apparatus according to Embodiment 1 of the present invention. In the following description, FIGS. Moreover, in this Embodiment 1, the action determination method is implemented by operating the action determination apparatus 10. Therefore, the description of the behavior determination method in the first embodiment is replaced with the following description of the operation of the behavior determination device 10. In addition, the following description is performed about the case where the action determination apparatus 10 is constructed | assembled by the server computer as shown in FIG. However, the same applies to the case where the behavior determination device 10 is constructed by a terminal device or a computer other than the server computer.

  First, when the data acquisition unit 3 of the terminal device 1 outputs the time-series sensor data output from the acceleration sensor to the time window extraction unit 11, as shown in FIG. Data is acquired (step A1). For example, when a person to be measured carries a mobile phone with a built-in acceleration sensor and the data acquisition unit 3 is a part of the mobile phone, the sensor data is acquired through the mobile phone and the behavior determination device (server computer) 10 Are sent to the time window extraction unit 11. In step A1, the sensor data acquired by the time window extraction unit 11 is digital data, and is composed of a set of output values at each data point. Data points are acquired at regular intervals.

  Next, the time window cutout unit 11 cuts out the acquired sensor data for a predetermined time length such as 1 second, for example, and sends the cut out sensor data to the peak extraction unit 13 as time window data. Output (step A2).

  Next, the peak extraction unit 13 uses the time window data output from the time window extraction unit 12 to extract peak information that identifies a peak existing in the time window data (step A3). In Step A3, the peak extraction unit 13 calculates a feature amount other than the floating degree, for example, an acceleration dispersion value and a peak interval.

  Subsequently, the peak extraction unit 13 determines from the peak information extracted in step A3 whether or not a peak corresponding to the ground contact of one step exists in the time window data (step A4). If the result of determination in step A4 does not exist, steps A2 and A3 are executed again.

  On the other hand, if it exists as a result of the determination in step A4, the peak extraction unit 13 outputs peak information to the peak-to-peak data extraction unit 14, and the peak-to-peak data extraction unit 14 executes step A5. At this time, the peak extraction unit 13 outputs the calculated feature amount to the behavior determination unit 16.

  In step A5, the peak-to-peak data cutout unit 14 identifies the latest peak and the peak immediately preceding it based on the peak information output from the peak extraction unit 13, and the time zone corresponding to these peaks. Cut out sensor data. Further, the peak-to-peak data extraction unit 14 outputs the extracted peak-to-peak data to the floating degree calculation unit 15.

  Next, the floating degree calculation unit 15 calculates the floating degree using the peak-to-peak data output from the peak-to-peak data cutout unit 14 (step A6). For example, when the peak-to-peak depth area is used as the floating degree, the floating-degree calculating unit 15 integrates a portion that is equal to or less than the set value of the peak-to-peak data with time, and calculates the peak-to-peak depth area.

  Specifically, in the first embodiment, since the sensor data is digital data as described above, the floating degree calculation unit 15 performs integration by the following processing. First, the buoyancy degree calculation unit 15 obtains the difference between 1G and the acceleration value of the data point for each data point having a value of 1G or less of the peak-to-peak data, and calculates the sum of the whole difference value from the obtained 1G. calculate. The calculated sum corresponds to the peak-to-peak depth area obtained by integration. Further, the floating degree calculation unit 15 outputs the calculated floating degree to the behavior determination unit 16.

  Next, the behavior determination unit 16 performs a behavior determination using the time window feature amount other than the floating degree output from the peak extraction unit 13 and the floating degree output from the floating degree calculation unit 15 (step S7). ). For example, when the peak extraction unit 13 calculates the acceleration dispersion value and the peak interval as the feature amounts, the behavior determination unit 16 determines whether the determination rule shown in FIGS. 4 to 6 or FIGS. The behavior of the measurement target is determined using the indicated determination rule. Further, the behavior determination unit 16 outputs the determination result to the output unit 17.

  Next, the output part 17 transmits the determination result output from the action determination part 16 to the terminal device 1 (step A8). When step A8 is executed, in the terminal device 1, the output device (display device) 4 displays the transmitted determination result on the screen as, for example, character information or image information. Although not shown in FIG. 3, the output unit 17 may output the determination result to either or both of the storage device of the behavior determination device 10 and the storage device of the terminal device 1.

  The processing in steps A1 to A8 described above is performed at regular intervals, such as once per second, or every time acceleration sensor data is newly output from the data acquisition unit 3, etc. Will be executed repeatedly. The user of the terminal device 1 can obtain a new determination result every time the processing of steps A1 to A8 is executed in the behavior determination device 10.

  Moreover, the program in Embodiment 1 of this invention should just be a program which makes computers, such as a server computer or a personal computer, perform step A1-A8 shown in FIG. By installing and executing this program on a computer, the behavior determination device 10 and the behavior determination method according to the first embodiment can be realized. In this case, a CPU (Central Processing Unit) of the computer functions as the time window extraction unit 11, the feature amount calculation unit 12, and the behavior determination unit 16, and performs processing. The communication interface of the computer functions as the output unit 17. A specific configuration of the computer will be described later.

  Here, the effect of the first embodiment of the present invention will be described. According to the first embodiment, it is possible to perform behavior determination using various feature amounts calculated from the time window data, including the degree of floating. In particular, the degree of buoyancy is an index that characterizes the vertical movement of the entire body that appears when the measurement target moves in the vertical direction. In the first embodiment, since the behavior determination is performed using the degree of float that is such an index, the vertical movement represented by the stair ascending / descending is performed regardless of how the acceleration sensor is held in the measurement target. The accuracy can be determined.

(Embodiment 2)
Next, a behavior determination device, a behavior determination system, a terminal device, a behavior determination method, and a program according to Embodiment 2 of the present invention will be described with reference to FIGS. Initially, the structure of the action determination apparatus and action determination system in this Embodiment 2 is demonstrated using FIG. FIG. 12 is a block diagram illustrating a configuration of the behavior determination device and the behavior determination system according to Embodiment 2 of the present invention.

  As shown in FIG. 12, the behavior determination apparatus 20 in the second embodiment includes a normalization processing unit 18 in the feature amount calculation unit 12, and in this respect, the behavior in the first embodiment shown in FIG. This is different from the determination device 10. About the point other than this, the action determination apparatus 20 is comprised similarly to the action determination apparatus 10. FIG. The behavior determination system 102 according to the second embodiment includes the terminal device 1 and the behavior determination device 20 that are also illustrated in FIG. Hereinafter, the difference from the first embodiment will be mainly described.

  The normalization processing unit 18 performs normalization on the floating degree calculated by the floating degree calculating unit 15 using a feature amount other than the floating degree. Unlike the first embodiment, the behavior determination unit 16 performs behavior determination using the normalized floating degree. In the second embodiment, normalization is performed in order to correct a difference in output of sensor data caused by a difference in how the acceleration sensor is held for each measurement target, a pace, and the like.

  In the second embodiment, as a specific method of normalization, for example, when the peak-to-peak depth area is used as the degree of floatation, the peak interval value calculated by the peak extraction unit 13 is used. There is a method of dividing the inter-depth area. In this case, the normalization processing unit 18 acquires the calculated peak interval from the peak extraction unit 13 and uses it for the normalization processing. In addition, the normalization processing unit 18 can acquire only peak information from the peak extraction unit 13, calculate a peak interval from the peak information, and perform normalization processing using the calculated peak interval. The normalization processing unit 18 outputs the value (normalized floating degree) obtained by such normalization processing to the behavior determination unit 16.

  In the second embodiment, the specific normalization method is not limited to the method described above. This will be described below. For example, when the person to be measured wears the acceleration sensor 2 on the trunk, the peak due to the grounding of the foot is almost the same peak in the sensor data regardless of whether the left or right foot is grounded. Appears in shape. On the other hand, when the person to be measured wears the acceleration sensor 2 at a position other than the trunk, the peak that appears in the sensor data depends on which side the left side of the acceleration sensor 2 is worn. The shape changes. That is, the peak shape that occurs when the right foot touches is different from the peak shape that occurs when the left foot touches. The “trunk” as used in this specification refers to the region from the bottom of the neck to the base of the foot, and particularly near the center of gravity of the body.

  Therefore, normalization can also be performed using, for example, the peak value obtained by the peak extraction unit 13 or the length of a section of 1 G or less between peaks (see FIG. 2) in addition to the peak interval. In addition, normalization can be performed using a value such as an acceleration dispersion value calculated by the peak extraction unit 13. In addition, when performing such a normalization process, the normalization process part 18 accesses the peak extraction part 13 as needed, and acquires required information. The normalization processing unit 18 can also use a combination of these normalization methods described above.

  Moreover, in this Embodiment 2, since the action determination part 16 performs action determination based on the normalized floating degree, for example, the determination rule shown in FIG. 13 is used. FIG. 13 is a diagram illustrating an example of a determination rule used in Embodiment 2 of the present invention. In the example of FIG. 13, the peak-to-peak depth area shown in FIG. 2 is used as the degree of floating, and a determination rule when the peak depth area is normalized by the peak interval is shown. When the behavior determining unit 16 obtains the normalized peak depth area, it applies this to the determination rule shown in FIG. 13, and the behavior to be measured is “walking (moving on flat ground)”, “rising stairs” or “ It is determined whether it falls under “step down”.

  Next, operation | movement of the action determination apparatus 20 in Embodiment 2 of this invention is demonstrated using FIG. FIG. 14 is a flowchart showing the operation of the behavior determination apparatus according to Embodiment 2 of the present invention. In the following description, FIG. 12 is taken into consideration as appropriate. In the second embodiment, the behavior determination method is implemented by operating the behavior determination device 20. Therefore, the description of the behavior determination method in the second embodiment is replaced with the following description of the operation of the behavior determination device 20. In addition, the following description is performed about the case where the action determination apparatus 20 is constructed | assembled by the server computer as shown in FIG. However, the same applies to the case where the behavior determination device 20 is constructed by a computer other than the terminal device or the server computer.

  As shown in FIG. 14, first, Step A11 to Step A16 are executed by the time window extraction unit 11 and the feature amount calculation unit 12 to calculate the floating degree. Steps A11 to A16 shown in FIG. 14 are the same steps as Steps A1 to A6 shown in FIG.

  In the second embodiment, unlike the first embodiment, the floating degree calculation unit 15 outputs the calculated value to the normalization processing unit 18 instead of the behavior determination unit 16. In addition, the peak extraction unit 13 outputs the calculated feature amount to the normalization processing unit 18 in addition to the behavior determination unit 16.

  Next, the normalization processing unit 18 performs normalization processing on the given floating degree (step A17). Specifically, for example, assuming that the peak-to-peak depth area is used as the floating degree, in step A17, the normalization processing unit 18 acquires the peak interval from the peak extracting unit 13, and uses this value to determine the peak-to-peak range. The normalization process is performed by dividing the depth area. Further, the normalization processing unit 18 may acquire only peak information from the peak extraction unit 13 and divide the peak-to-peak depth area by the peak interval value calculated based on the peak information to perform normalization processing. . Thereafter, the normalization processing unit 18 outputs the normalized floating degree to the behavior determination unit 16.

  Next, the behavior determination unit 16 performs a behavior determination using the time window feature amount other than the floating degree output from the peak extracting unit 13 and the floating degree normalized by the normalization processing unit 18 (step) S18). Specifically, the behavior determination unit 16 uses the determination rule shown in FIG. 13 to determine whether the measurement target behavior is “walking (moving on a flat ground)”, “ascending stairs”, or “descending stairs”. Judge whether to do. For example, if the normalized value of the floating degree is, for example, 208 [mG], the behavior determination unit 16 can determine that the behavior at this time is descending the stairs.

  Also in the second embodiment, for example, when the peak extraction unit 13 calculates the acceleration dispersion value and the peak interval as the feature amount, the behavior determination unit 16 is illustrated in FIGS. 4 and 5. The determination using the determination rule or the determination rule shown in FIGS. 7 and 8 can also be performed.

  Thereafter, when the behavior determination unit 16 outputs the determination result to the output unit 17, the output unit 17 transmits the determination result output from the behavior determination unit 16 to the terminal device 1 (step A19). Step A19 is the same as step A8 shown in FIG.

  The processing of the above steps A11 to A19 is performed at regular intervals such as once per second, or every time acceleration sensor data is newly output from the data acquisition unit 3, etc. Is executed repeatedly. The user of the terminal device 1 can obtain a new determination result every time the processing of steps A11 to A19 is executed in the behavior determination device 20.

  Moreover, the program in Embodiment 2 of this invention should just be a program which makes computers, such as a server computer or a personal computer, perform step A11-A19 shown in FIG. By installing and executing this program on a computer, the behavior determination device 20 and the behavior determination method according to the second embodiment can be realized. In this case, a CPU (Central Processing Unit) of the computer functions as the time window extraction unit 11, the feature amount calculation unit 12, the behavior determination unit 16, and the normalization processing unit 18 to perform processing. The communication interface of the computer functions as the output unit 17. A specific configuration of the computer will be described later.

  Here, the effect of the second embodiment of the present invention will be described. For example, when the peak-to-peak depth area is used as the degree of floatation, the value of the peak-to-peak depth area has a property that the value increases when the pace is slow, and the value decreases when the pace is fast. The degree of floating is also affected by the pace. Therefore, an erroneous determination may occur when the pace changes. On the other hand, if the obtained peak-to-peak depth area is divided by the peak interval representing the pace, and the value is used as the feature amount, the influence due to the change in the pace can be removed. That is, according to the second embodiment, it is possible to perform the action determination using the value obtained by normalizing the floating degree, and thus it is possible to perform the action determination with higher accuracy without being affected by the pace.

(Embodiment 3)
Next, a behavior determination device, a behavior determination system, a terminal device, a behavior determination method, and a program according to Embodiment 3 of the present invention will be described with reference to FIGS. 15 and 16. Initially, the structure of the action determination apparatus and action determination system in this Embodiment 3 is demonstrated using FIG. FIG. 15 is a block diagram illustrating a configuration of the behavior determination device and the behavior determination system according to Embodiment 3 of the present invention.

  As shown in FIG. 15, the action determination device 30 in the third embodiment includes a history storage unit 31, and is different in this respect from the action determination device 10 in the first embodiment shown in FIG. 3. . In addition, since the history storage unit 31 is provided, the processing in the behavior determination unit 16 is also different from that in the first embodiment. About the point other than these, the action determination apparatus 30 is comprised similarly to the action determination apparatus 10. FIG. The action determination system 103 according to the third embodiment includes the terminal device 1 and the action determination device 30 that are also shown in FIG. Hereinafter, the difference from the first embodiment will be mainly described.

  In the first embodiment, the floating degree calculating unit 15 outputs the calculated floating degree directly to the action determining unit 16 (see FIG. 3). However, in the third embodiment, the floating degree calculating unit 15 First, the floating degree calculated by is stored in the history storage unit 31.

  The history storage unit 31 calculates the feature amount calculated by the feature amount calculation unit 12, specifically, the float degree calculated by the float degree calculation unit 15 and the feature amount calculated by the peak extraction unit 13 during the set period. ,Remember. In addition, the history storage unit 31 may store peak information for a certain past time. Hereinafter, the various feature quantities including the floating degree stored in the history storage unit 31 are referred to as “history information”.

  Further, when the latest float degree is given from the float degree calculating unit 15 or another latest feature amount is given from the peak extracting unit 13, the history storage unit 31 stores the given latest value. Then, the history storage unit 31 updates the history information as necessary, for example, by erasing the oldest history information among the history information already stored.

  In the third embodiment, the behavior determination unit 16 accesses the history storage unit 31, acquires the latest history information stored therein, and performs the behavior determination of the measurement target using the acquired history information. . At that time, the behavior determination unit 16 can obtain a standard value of the feature quantity including the degree of float using the history information, and can perform behavior judgment using the obtained standard value.

  Here, as an example, an acceleration dispersion value, a peak interval, and a floating degree are considered as feature amounts used for action determination. In this case, the behavior determination unit 16 calculates an average value for each past fixed time as these standard values. And the action determination part 16 can apply the calculated average value to the determination rule shown in FIGS. 4-6, and can perform the action determination of a measuring object.

  Specifically, for example, it is assumed that the peak-to-peak depth area is calculated as the degree of floating, and the history storage unit 31 stores the latest three peak-to-peak depth areas as the history information. In this case, the behavior determination unit 16 obtains an average value of these latest three values, and performs behavior determination using the average value. If the history of the actual peak-to-peak depth area is 87 [mG · sec], 97 [mG · sec], and 101 [mG · sec], the behavior determination unit 16 determines the depth area between these peaks. Find the average value. The average value is 95 [mG · sec]. When the action determination unit 16 applies the obtained average value to the determination rule shown in FIG. 6, the action to be measured is determined to be “step down”. The behavior determination unit 16 outputs the determination result to the output unit 7 also in the third embodiment.

  In the third embodiment, the standard value is not limited to the average value calculated from the history information. Furthermore, the use of the standard value obtained from the history information is not limited to the use described above. For example, the behavior determination unit 16 may use the obtained standard value, consider an extremely large value or an extremely small value as noise compared to the standard value, and perform processing to remove these. When such processing is performed, the accuracy of the action determination result is further improved.

  In addition, the behavior determination unit 16 calculates an average value for the feature amount excluding the latest feature amount among the feature amounts included in the history information, and obtains a difference between the calculated average value and the latest feature amount, When the obtained difference value is equal to or less than the threshold value, the determination of the action to be measured can be stopped. In this case, the threshold value of the difference includes 0 (zero) or a value close to 0 (zero).

  When processing using such a difference is performed, unnecessary processing is not executed, and thus the burden on the behavior determination device 30 is reduced. Further, such processing is particularly useful when the behavior determination device 30 is incorporated in the terminal device 1 such as a mobile phone that has large restrictions on calculation resources and power consumption (see FIG. 10).

  Next, operation | movement of the action determination apparatus 30 in Embodiment 3 of this invention is demonstrated using FIG. FIG. 16 is a flowchart showing the operation of the behavior determination apparatus according to Embodiment 3 of the present invention. In the following description, FIG. 15 is referred to as appropriate. In the third embodiment, the behavior determination method is implemented by operating the behavior determination device 30. Therefore, the description of the behavior determination method in the third embodiment is replaced with the following description of the operation of the behavior determination device 30. In addition, the following description is performed about the case where the action determination apparatus 30 is constructed | assembled by the server computer as shown in FIG. However, the same applies to the case where the behavior determination device 30 is constructed by a computer other than the terminal device or the server computer.

  As shown in FIG. 16, first, Step A21 to Step A26 are executed by the time window cutout unit 11 and the feature amount calculation unit 12 to calculate the floating degree. Steps A21 to A26 shown in FIG. 16 are the same as steps A1 to A6 shown in FIG.

  In the third embodiment, unlike the first embodiment, the floating degree calculation unit 15 and the peak extraction unit 13 output the calculated values to the history storage unit 31 instead of the behavior determination unit 16.

  Next, when the history storage unit 31 receives the calculated feature amount from the floating degree calculation unit 15 or the peak extraction unit 13, it stores it as the latest history information and further updates the history information (step A27). . Specifically, in step A27, the history storage unit 31 is given when the latest float degree is given from the float degree calculating unit 15 or the latest other feature amount is given from the peak extracting unit 13. Remember the latest value. Then, the history storage unit 31 updates the history information as necessary, for example, by erasing the oldest history information among the history information already stored.

  Next, the behavior determination unit 16 accesses the history storage unit 31, acquires the latest history information, and performs a behavior determination using this (step A28). Specifically, the behavior determination unit 16 acquires, for example, history information on the acceleration dispersion value, the peak interval, and the floating degree, and calculates an average value for these past certain time periods. And the action determination part 16 applies each calculated average value to the determination rule shown in FIGS. 4-6, and performs the action determination of a measuring object.

  Thereafter, when the behavior determination unit 16 outputs the determination result to the output unit 17, the output unit 17 transmits the determination result output from the behavior determination unit 16 to the terminal device 1 (step A29). Step A29 is the same as step A8 shown in FIG.

  The processing of the above steps A21 to A29 is performed at regular intervals, such as once per second, or every time acceleration sensor data is newly output from the data acquisition unit 3, etc. Will be executed repeatedly. The user of the terminal device 1 can obtain a new determination result every time the processing of steps A21 to A29 is executed in the behavior determination device 30.

  Moreover, the program in Embodiment 3 of this invention should just be a program which makes computers, such as a server computer or a personal computer, perform step A21-A29 shown in FIG. By installing and executing this program on a computer, the behavior determination device 30 and the behavior determination method according to the third embodiment can be realized. In this case, a CPU (Central Processing Unit) of the computer functions as the time window extraction unit 11, the feature amount calculation unit 12, the behavior determination unit 16, and the normalization processing unit 18 to perform processing. The communication interface of the computer functions as the output unit 17. Further, a storage device provided in the computer functions as the history storage unit 31. A specific configuration of the computer will be described later.

  Here, the effect of the third embodiment of the present invention will be described. In the third embodiment, a value such as an average value is calculated based on past history information, and behavior determination is performed based on this value. For this reason, even when noise is suddenly mixed in the sensor data from the acceleration sensor, occurrence of a situation in which the value of the feature value used for action determination is extremely changed by the sudden noise is suppressed. As a result, according to the third embodiment, the occurrence of erroneous determination due to sudden noise can be suppressed, and as a result, more accurate action determination can be performed.

  The third embodiment can be combined with the second embodiment. Specifically, the behavior determination device 30 illustrated in FIG. 15 may include the normalization processing unit 18 illustrated in FIG. In this case, in the behavior determination device 30, the history storage unit 31 stores the normalized floating degree as history information.

(Embodiment 4)
Next, a behavior determination device, a behavior determination system, a terminal device, a behavior determination method, and a program according to Embodiment 4 of the present invention will be described with reference to FIGS. Initially, the structure of the action determination apparatus and action determination system in this Embodiment 4 is demonstrated using FIG. FIG. 17 is a block diagram showing the configuration of the behavior determination device and the behavior determination system according to Embodiment 4 of the present invention.

  As shown in FIG. 17, the behavior determination device 40 according to the fourth embodiment includes a learning unit 41, and is different from the behavior determination device 10 according to the first embodiment shown in FIG. 3 in this respect. In addition, since the learning unit 41 is provided, the processing in the behavior determination unit 16 is also different from that in the first embodiment. About the point other than these, the action determination apparatus 40 is comprised similarly to the action determination apparatus 10. FIG.

  Moreover, the action determination system 104 in this Embodiment 4 is provided with the terminal device 1 and the action determination apparatus 40 which were also shown in FIG. In the terminal device shown in FIG. 17, “input device 5” not shown in FIGS. 3, 10, 12, and 15 is shown. Hereinafter, the difference from the first embodiment will be mainly described.

  The learning unit 41 is provided for performing behavior determination in consideration of individual differences between persons to be measured. In the fourth embodiment, when the output unit 17 outputs the determination result to the terminal device 1 and the output device 4 displays the determination result, the user who knows the correct action uses the input device 5 to determine whether the determination result is correct. You can enter the correct answer. As a result, the terminal device 1 outputs information (hereinafter referred to as “correct answer behavior information”) that specifies the correctness of the determination result to the behavior determination device 40.

  For example, if the terminal device 1 is a mobile phone, a person (user) to be measured first knows the determination result of his / her action through the mobile phone. Specifically, the screen shown in FIG. 18 is displayed on the screen of the output device (display device) 4. FIG. 18 is a diagram illustrating an example of a screen of the terminal device when the user is requested to input whether the determination result is correct in Embodiment 4 of the present invention.

  When the determination result is correct, the user inputs that the determination result is correct using a key of the mobile phone or a touch panel. On the other hand, if the determination result is incorrect, the user inputs the correct action using the key of the mobile phone or the touch panel. Thereby, the correct behavior information corresponding to the user input result is transmitted from the terminal device 1 to the behavior determination device 40.

  When the terminal device 1 transmits the correct behavior information, the learning unit 41 receives the correct behavior information, compares the received correct behavior information with the determination result transmitted by the output unit 17 earlier, and from the comparison result, The rules when the behavior determination unit 16 performs the determination are updated. In the fourth embodiment, the learning unit 41 sets the threshold value set in the determination rule shown in FIGS. 4 to 6 or the determination rule shown in FIGS. 7 to 9 based on the received correct behavior information. Make adjustments.

  For example, when the peak-to-peak depth area is used as the floating degree and the value of the peak-to-peak depth area calculated by the floating degree calculating unit 15 is 110 [mG · sec], the determination rule illustrated in FIG. The action determination unit 16 determines “step down”. In this case, if the user inputs “step up” as correct action information, the learning unit 41 performs learning and updates the determination rule illustrated in FIG. 6 to the determination rule illustrated in FIG. Thereafter, the behavior determination unit 16 executes determination processing using the updated determination rule. FIG. 19 is a diagram illustrating an example of a determination rule changed after learning in the fourth embodiment of the present invention.

  Next, operation | movement of the action determination apparatus 40 in Embodiment 4 of this invention is demonstrated using FIG. FIG. 20 is a flowchart showing the operation of the behavior determination apparatus according to Embodiment 4 of the present invention. In the following description, FIG. 17 is referred to as appropriate. Moreover, in this Embodiment 4, the action determination method is implemented by operating the action determination apparatus 40. FIG. Therefore, the description of the behavior determination method in the fourth embodiment is replaced with the following description of the operation of the behavior determination device 40. In addition, the following description is performed about the case where the action determination apparatus 40 is constructed | assembled by the server computer as shown in FIG. However, the same applies to the case where the behavior determination device 40 is constructed by a computer other than the terminal device or the server computer.

  As shown in FIG. 20, first, Step A31 to Step A36 are executed by the time window cutout unit 11 and the feature amount calculation unit 12 to calculate the floating degree. Steps A31 to A36 shown in FIG. 20 are the same steps as steps A1 to A6 shown in FIG.

  Next, the behavior determination unit 16 performs a behavior determination using the time window feature amount other than the floating degree output from the peak extraction unit 13 and the floating degree output from the floating degree calculation unit 15 (step S37). ). Step A37 is the same as step A7 shown in FIG. 11. However, when the determination rule has already been changed by the learning unit 41, the behavior determination unit 16 determines that the determination rule after the change (FIG. 19). Further, the behavior determination unit 16 outputs the result of the behavior determination to the output unit 17.

  Next, the output part 17 transmits the determination result output from the action determination part 16 to the terminal device 1 (step A38). When step A38 is executed, in the terminal device 1, the output device (display device) 4 displays the transmitted determination result on the screen.

  However, in step A38, unlike step A8 shown in FIG. 11, the screen of the terminal device 1 displays a screen prompting the user to input whether the determination result is correct or not, as shown in FIG. The user inputs a correct answer when the determination result is correct, and selects and inputs a correct action result when the determination result is not correct.

  Next, after the transmission by the output unit 17, the learning unit 41 is in a standby state until the correct behavior information is transmitted from the terminal device 1, and when there is a transmission, accepts the correct behavior information (step A39). ).

  Next, the learning unit 41 determines whether or not the determination result in Step A37 is correct from the correct behavior information (Step A40). As a result of the determination, if the answer is correct, the learning unit 41 ends the process without updating the determination rule. On the other hand, if the result of determination is not correct, the learning unit 41 performs learning based on the correct behavior information and updates the determination rule (step A41).

  After execution of step A41, the behavior determination unit 16 performs the determination process again using the updated determination rule (step A37). Subsequently, steps A38 to A40 are executed again. Steps A37 to A40 and A41 are repeated until the user inputs that the answer is correct.

  The processing of the above steps A31 to A41 is performed at regular intervals, such as once per second, or every time acceleration sensor data is newly output from the data acquisition unit 3, etc. Is executed repeatedly. The user of the terminal device 1 can obtain a new determination result every time the processing of steps A31 to A41 is executed in the behavior determination device 40.

  Moreover, the program in Embodiment 4 of this invention should just be a program which makes a computer, such as a server computer or a personal computer, perform step A31-A41 shown in FIG. By installing and executing this program on a computer, the behavior determination device 40 and the behavior determination method according to the fourth embodiment can be realized. In this case, a CPU (Central Processing Unit) of the computer functions as the time window extraction unit 11, the feature amount calculation unit 12, the behavior determination unit 16, and the learning unit 41 to perform processing. The communication interface of the computer functions as the output unit 17. A specific configuration of the computer will be described later.

  Here, the effect of the fourth embodiment of the present invention will be described. Generally, there is a difference in the magnitude of vertical movement of the body due to reasons such as different heights. For this reason, depending on the physical characteristics of the person to be measured, if a threshold value set in the determination rule is used, the action may not be accurately determined. However, in the fourth embodiment, the determination rule is updated based on the correct action declaration from the person who knows the correct action, specifically, the threshold value is adjusted. Therefore, according to the fourth embodiment, it is possible to set an optimum threshold value adapted to each individual by repeating learning based on correct behavior, so that any person can be accurately acted on. It becomes possible to judge.

  Further, in the example of the threshold value change when an erroneous determination occurs described with reference to FIG. 19, the value of the peak-to-peak depth area when the erroneous determination occurs is adopted as a new threshold value, and the determination rule is updated. Has been done. However, in the fourth embodiment, the mode of threshold change is not limited to the above example.

  For example, as described above, 90 [mG · sec] is set as the upper limit threshold for determining the climb of the stairs (see FIG. 6), while the calculated peak-to-peak depth is 110 [mG · sec]. Consider the case where the input action is “up stairs” in spite of being there. In such a case, in the fourth embodiment, for example, the upper limit threshold value for determining the climb of the stairs can be changed from 91 [mG · sec] to 91 [mG · sec]. That is, in the fourth embodiment, it is possible to reduce the change width of one threshold value, and suppress the occurrence of an adverse effect that the threshold value changes abruptly every time a change is made and the determination accuracy becomes unstable. can do.

  Furthermore, Embodiment 4 can be combined with either or both of Embodiment 2 and Embodiment 3. In particular, when the fourth embodiment is combined with the third embodiment, the learning unit 41 can also update the determination rule using the history information stored in the history storage unit 31 (see FIG. 15). In this case, the determination rule can be updated by, for example, determining the distribution state of the peak-to-peak depth area during the stair climbing action when considering the stair climbing action. Specifically, the learning unit 41 identifies, from the history information, a range that includes, for example, 80% or more of the peak-to-peak depth area that appears during the stairs climbing action, and determines an upper limit and a lower limit of the identified range, respectively. Set as a new threshold.

(Embodiment 5)
Next, a behavior determination device, a behavior determination system, a terminal device, a behavior determination method, and a program according to Embodiment 5 of the present invention will be described with reference to FIGS. Initially, the structure of the action determination apparatus and action determination system in this Embodiment 5 is demonstrated using FIG. FIG. 21 is a block diagram illustrating the configuration of the behavior determination device and the behavior determination system according to the fifth embodiment of the present invention.

  As shown in FIG. 21, the action determination device 50 according to the fifth embodiment includes a history storage unit 51 and a learning unit 52. In this regard, the action determination device 10 according to the first embodiment shown in FIG. Is different. Further, since the history storage unit 51 and the learning unit 52 are provided, the processing in the behavior determination unit 16 is also different from that in the first embodiment. About the point other than these, the action determination apparatus 50 is comprised similarly to the action determination apparatus 10. FIG.

  The action determination system 105 according to the fifth embodiment includes the terminal device 1 and the action determination device 50 shown in FIG. In the terminal device shown in FIG. 21, “input device 5” not shown in FIGS. 3, 10, 12, and 15 is shown. Hereinafter, the difference from the first embodiment will be mainly described.

  In the fifth embodiment, the behavior determination unit 16 outputs the feature amount including the degree of float used for the determination and the determination result of the behavior at this time to the history storage unit 51 after performing the behavior determination. . The history storage unit 51 stores the output feature value, the determination result, and the time (determination time) at which the determination by the behavior determination unit 16 is performed as a set and stores it as history information. The history storage unit 51 stores history information for a predetermined period of time (for a set period), for example, a predetermined length such as one day, and the history information that has already been stored as necessary. The history information is updated by, for example, sequentially deleting the oldest history information.

  In the fifth embodiment, the output unit 17 accesses the history storage unit 51 to acquire history information, and the determination result and the determination time for the set period stored in the history storage unit 51 are displayed. It transmits to the terminal device 1. Thereby, the output device 4 of the terminal device 1 displays the determination result and the determination time for the set period on the screen. Further, also in the fifth embodiment, when the output device 4 displays the output history information, the user who knows the correct behavior, as in the fourth embodiment, uses the input device 5 to determine whether the determination result is correct. You can enter the correct answer.

  Specifically, when history information is transmitted from the output unit 17, the terminal device 1 displays the determination time for the set period and the determination result at that time on the output device 4. The user of the terminal device 1 determines whether or not the displayed determination result correctly represents the true action at that time, and inputs correctness. As a result, the terminal device 1 outputs information (hereinafter referred to as “correct action information”) that specifies whether the determination result output by the output unit 17 is correct and the determination time to the behavior determination device 50.

  For example, when the terminal device 1 is a mobile phone, a person (user) to be measured first knows the determination result of his past behavior through the mobile phone. When the determination result is correct, the user inputs that the determination result is correct using a key of the mobile phone or a touch panel. On the other hand, if the determination result is incorrect, the user inputs the correct action using the key of the mobile phone or the touch panel. Thereby, the correct behavior information corresponding to the input result of the user is transmitted from the terminal device 1 to the behavior determination device 50.

  When the terminal device 1 transmits the correct action information, the learning unit 52 receives the correct action information, compares the received correct action information with the determination result and the determination time transmitted by the output unit 17 earlier, and compares From the result, the rule when the behavior determination unit 16 performs the determination is updated. Also in the fifth embodiment, as in the fourth embodiment, the learning unit 52 performs the determination rule illustrated in FIGS. 4 to 6 or illustrated in FIGS. 7 to 9 based on the accepted correct behavior information. In the determination rule, adjustment of a set threshold is performed.

  For example, when the peak-to-peak depth area is used as the floating degree and the value of the peak-to-peak depth area calculated by the floating degree calculating unit 15 is 110 [mG · sec], the determination rule illustrated in FIG. The action determination unit 16 determines “step down”. In this case, if the user inputs “step up” as correct action information, the learning unit 52 performs learning and updates the determination rule illustrated in FIG. 6 to the determination rule illustrated in FIG. Thereafter, the behavior determination unit 16 executes determination processing using the updated determination rule.

  Next, operation | movement of the action determination apparatus 50 in Embodiment 5 of this invention is demonstrated using FIG. FIG. 22 is a flowchart showing the operation of the behavior determination apparatus according to Embodiment 5 of the present invention. In the following description, FIG. 21 is taken into consideration as appropriate. In the fifth embodiment, the behavior determination method is implemented by operating the behavior determination device 50. Therefore, the description of the behavior determination method in the fifth embodiment is replaced with the following description of the operation of the behavior determination device 50. In addition, the following description is performed about the case where the action determination apparatus 50 is constructed | assembled by the server computer as shown in FIG. However, the same applies to the case where the behavior determination device 50 is constructed by a computer other than the terminal device or the server computer.

  As shown in FIG. 20, first, Step A51 to Step A56 are executed by the time window cutout unit 11 and the feature amount calculation unit 12 to calculate the floating degree. Steps A51 to A56 shown in FIG. 22 are the same steps as Steps A1 to A6 shown in FIG.

  Next, the behavior determination unit 16 performs a behavior determination using the time window feature amount other than the floating degree output from the peak extraction unit 13 and the floating degree output from the floating degree calculation unit 15 (step S57). ). In step S57, when the learning rule has already been changed by the learning unit 52, the behavior determination unit 16 makes a determination using the changed determination rule (see FIG. 19).

  Next, the behavior determination unit 16 outputs the feature amount including the degree of float used for the determination and the determination result to the history storage unit 51, and outputs the feature amount, the determination result, The determination time is set and stored as history information (step A58). At this time, the history storage unit 51 updates the history information as necessary.

  Next, the output unit 17 accesses the history storage unit 51 to acquire history information, and transmits the determination result and the determination time for the set period stored in the history storage unit 51 to the terminal device 1 ( Step A59). When step A59 is executed, in the terminal device 1, the output device (display device) 4 displays the transmitted determination result and determination time for the set period on the screen.

  In step A59, as in step A38 shown in FIG. 20, a screen for prompting the user to input the determination result and the determination time is displayed on the screen of the terminal device 1. The user inputs the correct answer when the determination result and the determination time are correct, and selects and inputs the correct action result when the determination result and the determination time are not correct.

  Next, after the transmission by the output unit 17 is performed, the learning unit 52 enters a standby state until the correct behavior information is transmitted from the terminal device 1, and when there is a transmission, accepts the correct behavior information (step A60). ).

  Next, the learning unit 52 determines whether or not all the determination results stored in the history storage unit 51 are correct from the correct behavior information (step A61). As a result of the determination, if all the answers are correct, the learning unit 52 ends the process without updating the determination rule. On the other hand, if even one answer is not correct as a result of the determination, the learning unit 52 performs learning based on the correct action information and updates the determination rule (step A62).

  After execution of step A62, the behavior determination unit 16 performs the determination process again using the history information and the updated determination rule (step A57). Subsequently, steps A58 to A61 are executed again. Steps A57 to A61 and A62 are repeated until the user inputs that the answer is correct.

  The processing of the above steps A51 to A62 is performed at regular intervals such as once per second, or every time acceleration sensor data is newly output from the data acquisition unit 3, etc. Is executed repeatedly. The user of the terminal device 1 can obtain a new determination result every time the processing of steps A51 to A62 is executed in the behavior determination device 50.

  Moreover, the program in Embodiment 5 of this invention should just be a program which makes a computer, such as a server computer or a personal computer, perform step A51-A62 shown in FIG. By installing and executing this program on a computer, the behavior determination device 50 and the behavior determination method in the fifth embodiment can be realized. In this case, a CPU (Central Processing Unit) of the computer functions as the time window extraction unit 11, the feature amount calculation unit 12, the behavior determination unit 16, and the learning unit 52 to perform processing. The communication interface of the computer functions as the output unit 17. Furthermore, a storage device provided in the computer functions as the history storage unit 51. A specific configuration of the computer will be described later.

  Here, the effect of the fifth embodiment of the present invention will be described. In the fifth embodiment, as in the fourth embodiment, it is possible to determine an optimum threshold value adapted to each individual by repeating learning based on correct behavior, so for any person It is possible to accurately determine the behavior.

  In particular, in the fifth embodiment, since the past actions for the set period of the person to be measured are displayed on the screen of the terminal device 1 at once, the person who knows the correct action can efficiently correct the action at an arbitrary timing. Can be entered. Therefore, according to the fifth embodiment, the person who knows the correct action does not need to input the correct incorrect answer every time the action is determined, and has an opportunity to input the correct incorrect answer at once for many determination results. Can be obtained. As a result, the convenience of inputting the correct / incorrect answer by the person who knows the correct action is improved, and as a result, the determination accuracy can be further improved.

  Further, in the fifth embodiment, as in the third embodiment, the behavior determination unit 16 may perform determination using the history information stored in the history storage unit 51. In this case, the fifth embodiment can obtain the same effects as those of the third embodiment.

  Here, the computer which implement | achieves an action determination apparatus by running the program in Embodiment 1-5 is demonstrated using FIG. FIG. 23 is a block diagram illustrating an example of a computer that can execute the program according to the first to fifth embodiments of the present invention.

  As shown in FIG. 23, the computer 110 includes a CPU 111, a main memory 112, a storage device 113, an input interface 114, a display controller 115, a data reader / writer 116, and a communication interface 117. These units are connected to each other via a bus 121 so that data communication is possible.

  The CPU 111 performs various calculations by developing the program (code) in the present embodiment stored in the storage device 113 in the main memory 112 and executing them in a predetermined order. The main memory 112 is typically a volatile storage device such as a DRAM (Dynamic Random Access Memory). Further, the program in the present embodiment is provided in a state of being stored in a computer-readable recording medium 120. Note that the program in the present embodiment may be distributed on the Internet connected via the communication interface 117.

  Specific examples of the storage device 113 include a hard disk and a semiconductor storage device such as a flash memory. The input interface 114 mediates data transmission between the CPU 111 and an input device 118 such as a keyboard and a mouse. The display controller 115 is connected to the display device 119 and controls display on the display device 119. The data reader / writer 116 mediates data transmission between the CPU 111 and the recording medium 120, and reads a program from the recording medium 120 and writes a processing result in the computer 110 to the recording medium 120. The communication interface 117 mediates data transmission between the CPU 111 and another computer.

  Specific examples of the recording medium 120 include general-purpose semiconductor storage devices such as CF (Compact Flash) and SD (Secure Digital), magnetic storage media such as a flexible disk, or CD-ROM (Compact Disk). Optical storage media such as Read Only Memory).

  When the behavior determination device is constructed by a terminal device such as a mobile phone, the program in the present embodiment is stored in the memory of the terminal device instead of the computer shown in FIG. Executed by.

  Hereinafter, the action determination device and the action determination system according to the first embodiment of the present invention will be described with reference to FIG. In addition, Example 1 corresponds to Embodiment 1, and in the following description, FIGS.

  In the first embodiment, an example will be described in which it is determined whether an action by a person to be measured is an action of “step up stairs”, “step down stairs”, or “walking (moving on a flat ground)”. Further, the person to be measured puts a mobile phone with a built-in acceleration sensor in the chest pocket, and the sensor data output by the acceleration is transmitted to the external action determination device 10 through the mobile phone. Further, the behavior determination device 10 determines the behavior of the person carrying the mobile phone using the sensor data obtained from the mobile phone, and returns the result to the mobile phone. The determination result is displayed on the screen of the mobile phone.

  An example of sensor data output by the acceleration sensor built in the mobile phone is shown in FIG. FIG. 24 is a diagram illustrating an example of sensor data output by the acceleration sensor in the first embodiment. The sensor data shown in FIG. 24 is sensor data obtained when a person actually goes down the stairs in a state where a mobile phone with a built-in acceleration sensor is put in a breast pocket. Is graphed. In the first embodiment, the sensor data when the person goes down the stairs shown in FIG. 24 is transmitted to the behavior determination apparatus 10 through the communication function of the mobile phone. Hereinafter, the operation in the behavior determination apparatus 10 will be described.

[Time 0 seconds-Time 1 second]
First, in the action determination device 10, the time window extraction unit 11 receives the sensor data transmitted from the data acquisition unit 3 of the terminal device 1. In the first embodiment, it is assumed that the time window cutout unit 11 cuts out sensor data so as to have an overlap period of 0.5 seconds between adjacent time windows in a length of 1 second.

  Here, consider a case where the time window extraction unit 11 extracts time window data for 1 second from time 0 seconds to time 1 seconds of the sensor data shown in FIG. The time window extraction unit 11 outputs the extracted time window data to the peak extraction unit 13.

  The peak extraction unit 13 extracts peaks using the time window data output from the time window extraction unit 11. Since the time window data is from the time 0 second to the time 1 second of the sensor data, there is one extracted peak, which exists at time 550 [msec]. Then, the peak extraction unit 13 outputs peak information for specifying the extracted peak to the peak-to-peak data extraction unit 14. Note that any of the methods described in the first embodiment is used as the peak extraction method, and a detailed description thereof is omitted here.

  The peak-to-peak data cutout unit 14 cuts out peak-to-peak data based on the given peak information. However, since it is immediately after the start of the determination process, the given peak information is one, and there are no two peaks that can be configured between the peaks. Therefore, as will be described later, time window data is cut out again by the time window cutting unit 11 (see step A4 shown in FIG. 11).

  In the first embodiment, the peak-to-peak data cannot be cut out. Therefore, the peak-to-peak data cutout unit 14 notifies the floating degree calculation unit 15 of the result that the peak-to-peak data cannot be cut out. The float degree calculation unit 15 is given the result that the peak-to-peak data cannot be cut out, and outputs the result that the float degree cannot be calculated to the behavior determination unit 16. Furthermore, since the behavior determination unit 16 is given a result that the degree of floating cannot be calculated, the behavior determination unit 16 sends a determination result indicating that the behavior cannot be determined or other behavior to the output unit 17, and the output unit 17 sends the result to the mobile phone. Send to. As a result, a determination result indicating that the determination is impossible or other behavior is displayed on the screen of the mobile phone.

[Time 0.5 seconds-Time 1.5 seconds]
Subsequently, the time window cutout unit 11 cuts out time window data for 1 second from time 0.5 seconds to time 1.5 seconds of the sensor data shown in FIG. The time window extraction unit 11 outputs the extracted time window data to the peak extraction unit 13.

  In this case as well, the peak extraction unit 13 extracts peaks using the time window data. In this case, two peaks are extracted from the time window data, the first being at 550 [msec] and the second being at time 1110 [msec]. The peak extraction unit 13 outputs the obtained peak information to the peak-to-peak data extraction unit 14.

  The peak-to-peak data extraction unit 14 extracts the peak-to-peak data in response to the occurrence times of the two peaks included in the given peak information being 550 [msec] and 1110 [msec]. . Of the two newly obtained peaks, the peak at the position of 550 [msec] is the same as the peak extracted in the previous peak extraction process, so the peak-to-peak data extraction unit 14 ignores the same peak. The peak generated at 1110 [msec] is taken as the latest peak.

  Then, the peak-to-peak data extraction unit 4 is between the latest peak (occurring at 1110 [msec]) and the immediately preceding peak (occurring at 550 [msec]) extracted by the previous peak extraction process. The sensor data is cut out from the sensor data acquired from the data acquisition unit 3. Further, the peak-to-peak data extraction unit 4 outputs the extracted sensor data to the floating degree calculation unit 15.

  The floating degree calculation unit 15 calculates the floating degree using the peak-to-peak data output from the peak-to-peak data cutout unit 14. In Example 1, since the peak-to-peak depth area of 1G or less is used as the degree of floating, the degree-of-floating calculation unit 15 specifies data having a value of 1G or less from the peak-to-peak data, and the acceleration The difference between the value and 1G is obtained, and the sum of the differences between the obtained data points is calculated. The calculated sum is the peak-to-peak depth area. Here, in Example 1, the calculated value of the peak-to-peak depth area (floating degree) was 104 [mG · sec]. The floating degree calculation unit 15 outputs the calculated value of the peak-to-peak depth to the behavior determination unit 16.

  Based on the peak-to-peak depth area output from the floating degree calculation unit 15, the behavior determination unit 16 determines whether the behavior to be measured is “up stairs”, “down stairs”, or “walking (moving on a flat ground)”. Is determined using the determination rule shown in FIG. Specifically, since the value of the depth area between the peaks is 104 [mG · sec], the behavior determination unit 16 determines that the behavior to be measured is “step down”. Thereafter, when the output unit 17 transmits the determination result to the mobile phone, the obtained determination result “down the staircase” is displayed on the screen of the mobile phone.

  The above operation is repeatedly executed at a predetermined interval, for example, once per second, and new determination results are sequentially displayed on the screen of the mobile phone each time a new determination result is obtained. At this time, the past determination history may be displayed together. As described above, according to the first embodiment, it is possible to accurately determine an action even when the measurement object moves in the vertical direction without being restricted with respect to how to attach the acceleration sensor.

  Next, the action determination apparatus and action determination system in Example 2 of this invention are demonstrated using FIG. In addition, Example 2 corresponds to Embodiment 2, and in the following description, FIGS. 12 to 14 are referred to as appropriate.

  Also in the second embodiment, as in the first embodiment, it is determined whether the action by the person to be measured is the action of “step climbing”, “step down”, or “walking (moving on flat ground)”. Will be described. However, in the second embodiment, unlike the first embodiment, it is assumed that the person to be measured puts the mobile phone with a built-in acceleration sensor in the right pocket of the pants.

  An example of data output by the acceleration sensor built in the mobile phone is shown in FIG. FIG. 25 is a diagram illustrating an example of sensor data output from the acceleration sensor in the second embodiment. The sensor data shown in FIG. 25 is acceleration sensor data obtained when a person actually goes down the stairs in a state where a mobile phone with a built-in acceleration sensor is put in the right pocket of the pants. Minutes are graphed. In Example 2, the sensor data shown in FIG. 25 when the person steps down the stairs is transmitted to the behavior determination device 20 through the communication function of the mobile phone. Hereinafter, the operation in the behavior determination apparatus 20 will be described.

  Also in the second embodiment, the time window extraction unit 11, the peak extraction unit 13, the peak-to-peak data extraction unit 14, and the floating degree calculation unit 15 operate in the same manner as in the first embodiment. Therefore, the description about the same operation is omitted.

  Also in the second embodiment, the time window extraction unit 11 performs time window data from time 0 seconds to time 1 seconds of sensor data and time window data from time 0.5 seconds to time 1.5 seconds of sensor data. And cut out. As a result, from the sensor data shown in FIG. 25, the peak c is extracted when the time is 520 [msec], and the peak d is extracted when the time is 1030 [msec].

  The floating degree calculation unit 15 uses a peak-to-peak depth area of 1 G or less as the floating degree as in the first embodiment. As a result, the value of the peak-to-peak depth area between the peak c and the peak d was 101 [mG · sec]. The floating degree calculation unit 15 outputs the calculated peak-to-peak depth to the normalization processing unit 18.

  The normalization processing unit 8 performs normalization processing of the calculated float degree. In the second embodiment, it is assumed that the value of the peak-to-peak depth area given as the floating degree is normalized by dividing by the value of the section length (see FIG. 2) of 1G or less. The normalization processing unit 18 given the peak-to-peak data from the peak-to-peak data cutout unit 14 obtains a section length of 1 G or less between peaks. As a result, in the sensor data shown in FIG. 25, the section length of 1G or less was 390 [msec] between 520 [msec] and 1030 [msec]. Therefore, the result of dividing the value of the peak-to-peak depth area given from the floating degree calculation unit 15 by the obtained section length of 1 G or less is 258 [mG]. The normalization processing unit 18 outputs this value to the behavior determination unit 16.

  The behavior determination unit 16 performs behavior determination using the determination rule shown in FIG. FIG. 26 is a diagram illustrating an example of a determination rule used in the second embodiment. In the determination rule shown in FIG. 26, a value normalized by dividing the peak-to-peak depth area by the section length of 1 G or less is used. For example, if the value calculated by normalization is 258 [mG], the behavior determination unit 16 uses the determination rule shown in FIG. 26, and the behavior at this time is “step down”. Is determined. And the action determination part 16 outputs the obtained determination result to the output part 17, and the output part 17 transmits a determination result to a mobile telephone. As a result, the determination result “step down” is displayed on the screen of the mobile phone.

  Further, the above operation is repeatedly executed at a predetermined interval, for example, once per second, and new determination results are sequentially displayed on the screen of the mobile phone each time a new determination result is obtained. At this time, the past determination history may be displayed together.

(Explanation of effects of Example 2)
Here, the effect of the second embodiment will be described in more detail below. In the sensor data shown in FIG. 25, the time window extraction unit 11 and the peak extraction unit 13 perform time window data from time 0.5 seconds to time 1.5 seconds and time 1 second to time 2 seconds. It is assumed that peak extraction is performed on the time window data. As a result, from the sensor data shown in FIG. 25, the peak d is extracted when the time is 1030 [msec], and the peak e is extracted when the time is 1670 [msec].

  Here, it is assumed that the value of the peak-to-peak depth area between the peak d and the peak e calculated by the floating degree calculation unit 15 is 54 [mG · sec]. The floating degree calculation unit 15 outputs the calculated value of the peak-to-peak depth area to the normalization processing unit 18.

  The normalization processing unit 18 performs a process of normalizing the given peak-to-peak depth area value by dividing the value by the section length value of 1G or less. First, the normalization processing unit 18 obtains a section length of 1 G or less between peaks. As a result, since the section length of 1G or less was 250 [msec], the result of dividing the value of the peak-to-peak depth area calculated by the floating degree calculation unit 15 by the obtained section length of 1G or less is 216 [mG]. The normalization processing unit 18 outputs this value to the behavior determination unit 16.

  Also in this case, the behavior determination unit 16 performs the behavior determination using the determination rule shown in FIG. Since the value calculated by the normalization processing unit 18 is 216 [mG], the behavior determination unit 16 determines that the behavior at this time is “step down”. And the action determination part 16 outputs the obtained determination result to the output part 17, and the output part 17 transmits a determination result to a mobile telephone. As a result, the determination result “step down” is displayed on the screen of the mobile phone.

  Here, the peak-to-peak depth obtained using the time window data from time 0 seconds to time 1 seconds and time window data from time 0.5 seconds to time 1.5 seconds shown in FIG. The area is referred to as area a. Similarly, a peak-to-peak depth region obtained by using time window data from time 0.5 second to time 1 second and time window data from time 1 second to time 2 seconds is referred to as region b. I will do it.

  As described above, the sensor data shown in FIG. 25 is acquired in a state where the mobile phone incorporating the acceleration sensor is placed in the right pocket of the pants of the person to be measured. Therefore, in FIG. 25, peak c and peak e are peaks that occur when the left foot located on the opposite side of the pocket in which the mobile phone incorporating the acceleration sensor is put in contact with the ground. On the other hand, the peak d is a peak that occurs when the right foot located on the same side as the pocket in which the mobile phone incorporating the acceleration sensor is placed contacts the ground.

  As shown in FIG. 25, the shape of a peak generated at the time of grounding is different between the foot on the pocket side in which the mobile phone incorporating the acceleration sensor is placed and the foot on the opposite side. For this reason, the value of the depth area between peaks differs greatly between the area | region a and the area | region b between adjacent peaks. Actually, the area of the region a is 101 [mG · sec], whereas the area of the region b is 54 [mG · sec].

  Thus, when the acceleration sensor is located in a part other than the trunk of a person or the like to be measured, the value of the peak-to-peak depth area is not constant. For this reason, if the behavior determination is performed using the value of the peak-to-peak depth area as it is, the determination accuracy may be lowered. On the other hand, according to the second embodiment, as described above, the normalization process is performed even when the value of the peak-to-peak depth area is not constant due to the difference in contact between the left and right feet due to the position of the acceleration sensor. By performing the above, it is possible to perform highly accurate determination without being affected by the position of the acceleration sensor.

  Next, an action determination device and an action determination system in Example 3 of the present invention will be described. In addition, Example 3 corresponds to Embodiment 3, and in the following description, FIGS. 15 and 16 are referred to as appropriate.

  Also in the third embodiment, as in the first embodiment, it is determined whether the action by the person to be measured is the action of “step up”, “step down”, or “walk (moving on flat ground)”. An example will be described. Also in the third embodiment, as in the first embodiment, it is assumed that a person to be measured has a mobile phone with a built-in acceleration sensor in a chest pocket. Furthermore, in the third embodiment, the time window extraction unit 11, the peak extraction unit 13, the peak-to-peak data extraction unit 14, and the float calculation unit 15 operate in the same manner as in the first embodiment. Therefore, the description about the same operation is omitted.

  In the third embodiment, the behavior determination device 30 includes a history storage unit 31, and the floating degree calculation unit 15 outputs the calculated floating degree (depth area between peaks of 1 G or less) to the history storage unit 31. . The history storage unit 31 stores the value of the peak-to-peak depth for a certain past time, and when the latest value of the peak-to-peak depth area is output from the floating degree calculation unit 15, the latest peak-to-peak depth. Store the area value. Then, the history storage unit 31 updates the history information, for example, by deleting the oldest peak depth area value in the history that has already been stored as necessary. The behavior determination unit 16 acquires updated history information and performs behavior determination.

  Now, it is assumed that the latest three peak-to-peak depth area values are stored as history information, and the action determination unit 16 has acquired these values. Assume that the depth areas between peaks actually stored are 87 [mG · sec], 97 [mG · sec], and 101 [mG · sec], respectively.

  In the third embodiment, the behavior determination unit 16 calculates an average value of the peak-to-peak depth areas included in the history information, and performs behavior determination using the average value. Since the peak-to-peak depth area included in the history information is 87 [mG · sec], 97 [mG · sec], and 101 [mG · sec] as described above, the average value of these history information is 95 [mG · sec]. Therefore, when the determination rule shown in FIG. 6 is used, the behavior determination unit 16 can determine that the step is “down the stairs”. And the action determination part 16 outputs the obtained determination result to the output part 17, and the output part 17 transmits a determination result to a mobile telephone. As a result, the determination result “step down” is displayed on the screen of the mobile phone.

  Further, the above operation is repeatedly executed at a predetermined interval, for example, once per second, and new determination results are sequentially displayed on the screen of the mobile phone each time a new determination result is obtained. At this time, the past determination history may be displayed together.

  As described above, according to the third embodiment, since the value of the peak-to-peak depth area is averaged using the history information, even if noise is partially mixed in the sensor data, an error due to noise is caused. The occurrence of determination is suppressed. For example, as described above, in Example 3, the value of the peak-to-peak area is 87 [mG · sec] for some reason when going down the stairs, but only this value is taken out. If the determination rule shown in FIG. 6 is applied, it is determined that “up of the stairs”.

  In general, in human daily behavior, it is considered that noise is often mixed into acceleration sensor data due to an unexpected event, for example, when the course is shifted due to passing each other. Therefore, as in the third embodiment, it is useful to perform the action determination using the past history from the viewpoint of suppressing the influence due to the mixing of these noises, and as a result, it is possible to suppress a decrease in the determination accuracy.

  Next, an action determination device and an action determination system in Example 4 of the present invention will be described. In addition, Example 4 corresponds to Embodiment 4, and FIGS. 17 to 20 are referred to as appropriate in the following description.

  Also in the fourth embodiment, as in the first embodiment, it is determined whether the action by the person to be measured is the action of “step up stairs”, “step down stairs”, or “walk (moving on flat ground)”. An example will be described. Also in the fourth embodiment, as in the first embodiment, it is assumed that a person to be measured has a mobile phone with a built-in acceleration sensor in a breast pocket. Further, in the fourth embodiment, the time window extraction unit 11, the peak extraction unit 13, the peak-to-peak data extraction unit 14, and the floating degree calculation unit 15 operate in the same manner as in the first embodiment. Therefore, the description about the same operation is omitted.

  Here, it is assumed that the floating degree calculation unit 15 calculates the peak-to-peak depth area as the floating degree. Then, it is assumed that the value of the peak-to-peak depth area actually calculated by the floating degree calculation unit 15 is 110 [mG · sec]. The floating degree calculation unit 15 outputs the calculated value of the peak-to-peak depth area to the behavior determination unit 16.

  The behavior determination unit 16 uses the determination rule of FIG. 6 based on the value of the peak-to-peak depth area output from the floating degree calculation unit 15 to determine whether the behavior to be measured is “up stairs” or “down stairs”. "Or" walking (moving on flat ground) ". Specifically, the behavior determination unit 16 applies the determination rule illustrated in FIG. 6 when the value of the peak-to-peak depth area is 110 [mG · sec], and the behavior is “step down”. Is determined.

  And the action determination part 16 outputs the obtained determination result to the output part 17, and the output part 17 transmits a determination result to a mobile telephone. As a result, the determination result “step down” is displayed on the screen of the mobile phone.

  Next, when the determination result is the same as the actual action, the user of the mobile phone inputs that the answer is correct. On the other hand, for example, when the determination result is “step down” while the execution action is “up stairs”, the mobile phone user inputs the correct action. The input correct behavior information is transmitted to the learning 41 of the behavior determination device 40 through a mobile phone.

  The learning unit 41 that has received the correct behavior information adjusts the threshold value of the determination rule illustrated in FIG. 6 based on the received correct behavior information. For example, as described above, it is assumed that the value of the depth area between peaks is 110 [mG · sec], and the determination result based on the determination rule shown in FIG. Then, if the correct behavior information input from the user of the mobile phone is “step up”, the learning unit 41 changes the determination rule shown in FIG. 6 to the determination rule shown in FIG. Thereafter, the behavior determination unit 16 performs the determination process again while referring to the changed new determination rule.

  As described above, according to the fourth embodiment, the threshold value is adjusted by learning even when an error occurs in the determination result due to a difference in the physique of the person to be measured. Judgment can be performed.

  Next, an action determination device and an action determination system in Example 5 of the present invention will be described. In addition, Example 5 corresponds to Embodiment 5, and in the following description, FIGS. 21 and 22 are referred to as appropriate.

  Also in the fifth embodiment, as in the first embodiment, it is determined whether the action by the person to be measured is the action of “step up”, “step down”, or “walk (moving on a flat ground)”. An example will be described. Also in the fifth embodiment, as in the first embodiment, it is assumed that a person to be measured has a mobile phone with a built-in acceleration sensor in a breast pocket. Further, in the fifth embodiment, the time window extraction unit 11, the peak extraction unit 13, the peak-to-peak data extraction unit 14, and the floating degree calculation unit 15 operate in the same manner as in the first embodiment. Therefore, the description about the same operation is omitted.

  Here, it is assumed that the floating degree calculation unit 15 calculates the peak-to-peak depth area as the floating degree. Then, it is assumed that the value of the peak-to-peak depth area actually calculated by the floating degree calculation unit 15 is 110 [mG · sec]. The floating degree calculation unit 15 outputs the calculated value of the peak-to-peak depth area to the behavior determination unit 16.

  The behavior determination unit 16 uses the determination rule of FIG. 6 based on the value of the peak-to-peak depth area output from the floating degree calculation unit 15 to determine whether the behavior to be measured is “up stairs” or “down stairs”. "Or" walking (moving on flat ground) ". Specifically, the behavior determination unit 16 applies the determination rule illustrated in FIG. 6 when the value of the peak-to-peak depth area is 110 [mG · sec], and the behavior is “step down”. Is determined.

  Then, the behavior determination unit 16 outputs the obtained determination result to the history storage unit 51. The history storage unit 51 determines the above-described determination result of “step down”, 110 [mG · sec] which is the value of the peak-to-peak depth area used for the determination, and the time when the determination result is output (determination Time) is stored as a set. The determination time is assumed to be 1:00 pm 0 seconds. The history storage unit 51 stores history information for a predetermined period of time (for a set period), for example, a predetermined length such as one day, and the history information that has already been stored as necessary. The history information is updated by, for example, sequentially deleting the oldest history information.

  The output unit 17 accesses the history storage unit 51 to acquire history information, and transmits the determination result and the determination time for the set period stored in the history storage unit 51 to the mobile phone. Thereby, the determination result and the determination time for the set period are displayed on the screen of the mobile phone.

  Next, the user of the mobile phone, for example, browses past determination results and inputs a correct / incorrect answer at an arbitrary time such as the next day of behavior determination. The user of the mobile phone, for example, sees the determination result that yesterday's 1:00 pm 0 second was “step down”, and the actual action was “step up” through the mobile phone. Enter that. The input correct behavior information is transmitted to learning 52 of the behavior determination device 50 through a mobile phone.

  The learning unit 52 adjusts the threshold value of the determination rule shown in FIG. 6 based on the transmitted correct behavior information, and updates the determination rule. For example, as described above, it is assumed that the value of the depth area between peaks is 110 [mG · sec], and the determination result based on the determination rule shown in FIG. Then, if the correct behavior information input from the user of the mobile phone is “step up”, the learning unit 52 changes the determination rule shown in FIG. 6 to the determination rule shown in FIG. Thereafter, the behavior determination unit 16 performs the determination process again while referring to the changed new determination rule.

  As described above, according to the fifth embodiment, the mobile phone user can efficiently input the correct action at an arbitrary timing. Therefore, it is not necessary to input the correct / incorrect answer every time the action is determined. It is possible to obtain an opportunity to input a correct answer / incorrect answer at a time for the determination result. As a result, the convenience of correct / incorrect answer input by the user of the mobile phone is improved, and as a result, the determination accuracy can be further improved.

  As described above, according to the present invention, for example, a behavior of a person when going out is determined, calorie consumption calculation according to the behavior of the person is performed with high accuracy, or health management is performed using an action record. can do. In particular, according to the present invention, since the movement of a person or the like in the vertical direction can be determined, the floor movement can be detected, and for example, it can be applied to indoor navigation.

DESCRIPTION OF SYMBOLS 1 Terminal device 2 Acceleration sensor 3 Data acquisition part 4 Output device 5 Input device 10 Behavior determination apparatus (Embodiment 1)
DESCRIPTION OF SYMBOLS 11 Time window extraction part 12 Feature-value calculation part 13 Peak extraction part 14 Inter-peak data extraction part 15 Floating degree calculation part 16 Action determination part 17 Output part 18 Normalization process part 20 Action determination apparatus (Embodiment 2)
30 Action determination device (Embodiment 3)
31 History storage unit 40 Action determination device (Embodiment 4)
41 learning unit 50 action determination device (Embodiment 5)
51 History Storage Unit 52 Learning Unit 100 Action Determination System (Embodiment 1)
101 terminal apparatus 102 action determination system (Embodiment 2)
103 Action determination system (Embodiment 3)
104 Action determination system (Embodiment 4)
105 Behavior determination system (Embodiment 5)
110 Computer 111 CPU
112 Main Memory 113 Storage Device 114 Input Interface 115 Display Controller 116 Data Reader / Writer 117 Communication Interface 118 Input Device 119 Display Device 120 Recording Medium 121 Bus

Claims (38)

An action determination device for determining a measurement target action based on time-series sensor data output by an acceleration sensor,
A time window extraction unit that extracts time window data of a set time length from the sensor data; and
A feature amount calculation unit that calculates a feature amount that serves as an index of behavior evaluation of the measurement target;
An action determination unit for determining an action of the measurement target based on the feature amount;
With
The feature amount calculation unit extracts peak information for specifying a peak included in the time window data from the time window data, specifies one peak and a peak immediately before the peak information based on the peak information, A portion existing between two specified peaks in the sensor data is cut out as peak-to-peak data, and from the peak-to-peak data, the vertical acceleration applied to the measurement object is equal to or less than a set value as the feature amount. The degree of occurrence indicating the degree of occurrence of the state is calculated, and the acceleration dispersion value for each time window data and the interval between the two specified peaks are calculated as feature quantities other than the degree of floating. And
The behavior determination unit stops, runs, the behavior of the measurement target is assumed in advance based on at least one of the floating degree, the acceleration dispersion value, and the interval between the two specified peaks. , walking, climbing stairs, and down the stairs, two or more any in either the appropriate actions selected from, or for one of these behaviors, or corresponds thereto, determining, and wherein An action determination device. The feature amount calculation unit integrates a portion of the peak-to-peak data that is equal to or less than the set value with time, and the obtained integrated value is set as the floating degree.
The behavior determination apparatus according to claim 1, wherein the behavior determination unit determines the behavior of the measurement target based on a threshold value of the integral value set for each supposed behavior of the measurement target.   The behavior determination apparatus according to claim 1, wherein the set value is set to a value of gravitational acceleration. The behavior determination unit determines whether the measurement target is stopped based on the acceleration dispersion value, and determines that the measurement target is not stopped, based on the interval between the two specified peaks, When determining whether the measurement target is running or walking, and determining that the measurement target is walking, based on the floating degree, determine whether the measurement target is going up the stairs or going down the stairs, The behavior determination apparatus according to claim 1 . The behavior determination unit calculates a first evaluation value indicating the possibility that the measurement target is stopped based on the acceleration dispersion value, and the measurement target is calculated based on an interval between the two specified peaks. Calculating a second evaluation value indicating the possibility of running or walking, and based on the degree of floating, the possibility that the measurement object is climbing the stairs or descending the stairs The behavior determination device according to claim 1 , wherein a third evaluation value is calculated. A normalization processing unit that normalizes the floating degree using a feature amount other than the floating degree;
The action determination unit, instead of the floating degree, by using a value calculated by normalization by the normalization processing unit determines the behavior of the measurement object, according to one of claims 1 to 5 Behavior determination device. The behavior determination apparatus according to claim 6 , wherein the normalization processing unit normalizes the floating degree using an interval between the specified two peaks. A history storage unit for storing the feature amount calculated by the feature amount calculation unit as history information for a set period;
The action determination unit, by using the history information, determining the standard values of the feature amount, using the standard values to determine the behavior of the measurement object, behavior of any of claims 1-7 Judgment device. The behavior determination device according to claim 8 , wherein the behavior determination unit calculates an average value of the feature amounts included in the history information, and uses the calculated average value as the standard value. The behavior determining unit calculates an average value for the feature amount excluding the latest feature amount among the feature amounts included in the history information, and further calculates a difference between the latest feature amount and the average value, The behavior determination apparatus according to claim 8 , wherein when the obtained difference value is equal to or less than a threshold value, determination of the behavior of the measurement target is stopped. An output unit for outputting a result of determination by the behavior determination unit to the outside; and
The input from the outside of the information specifying whether the determination result is correct is accepted, the received information is compared with the output determination result, and the action determination unit determines from the comparison result. And a learning unit that updates a determination rule when performing
The action determination unit, by using the rules in which the learning unit has updated, determines the behavior of the measurement object, behavior determining apparatus according to any one of claims 1-10. The feature amount calculated by the feature amount calculation unit, the determination result based on the feature amount by the behavior determination unit, and the time when the determination by the behavior determination unit is performed are set as history information for a set period. A history storage unit for storing;
An output unit that outputs the result of the determination and the time for the set period stored in the history storage unit;
Accepting input from the outside of the information specifying the correctness between the determination result and the time output, comparing the received information with the determination result and the time output, from the comparison result A learning unit that updates a rule when the behavior determination unit makes a determination,
The action determination unit, by using the rules in which the learning unit has updated, determines the behavior of the measurement object, behavior determining apparatus according to any one of claims 1-5. An acceleration sensor that outputs time-series sensor data according to the behavior of the measurement target; and an action determination device that determines the behavior of the measurement target based on the sensor data;
The behavior determination device is configured to extract, from the sensor data, a time window extraction unit that extracts time window data of a set time length, and a feature amount calculation unit that serves as an index for behavior evaluation of the measurement target. And an action determination unit that determines an action of the measurement target based on the feature amount,
The feature amount calculation unit extracts peak information for specifying a peak included in the time window data from the time window data, specifies one peak and a peak immediately before the peak information based on the peak information, A portion existing between two specified peaks in the sensor data is cut out as peak-to-peak data, and from the peak-to-peak data, the vertical acceleration applied to the measurement object is equal to or less than a set value as the feature amount. The degree of occurrence indicating the degree of occurrence of the state is calculated, and the acceleration dispersion value for each time window data and the interval between the two specified peaks are calculated as feature quantities other than the degree of floating. And
The behavior determination unit stops, runs, the behavior of the measurement target is assumed in advance based on at least one of the floating degree, the acceleration dispersion value, and the interval between the two specified peaks. , walking, climbing stairs, and down the stairs, two or more any in either the appropriate actions selected from, or for one of these behaviors, or corresponds thereto, determining, and wherein A behavior determination system. An acceleration sensor that outputs time-series sensor data according to the behavior of the measurement target;
A time window extraction unit that extracts time window data of a set time length from the sensor data; and
A feature amount calculation unit that calculates a feature amount that serves as an index of behavior evaluation of the measurement target;
An action determination unit for determining an action of the measurement target based on the feature amount;
With
The feature amount calculation unit extracts peak information for specifying a peak included in the time window data from the time window data, specifies one peak and a peak immediately before the peak information based on the peak information, A portion existing between two specified peaks in the sensor data is cut out as peak-to-peak data, and from the peak-to-peak data, the vertical acceleration applied to the measurement object is equal to or less than a set value as the feature amount. The degree of occurrence indicating the degree of occurrence of the state is calculated, and the acceleration dispersion value for each time window data and the interval between the two specified peaks are calculated as feature quantities other than the degree of floating. And
The behavior determination unit stops, runs, the behavior of the measurement target is assumed in advance based on at least one of the floating degree, the acceleration dispersion value, and the interval between the two specified peaks. , walking, climbing stairs, and down the stairs, two or more any in either the appropriate actions selected from, or for one of these behaviors, or corresponds thereto, determining, and wherein A terminal device. A method for determining an action to be measured based on time-series sensor data output from an acceleration sensor,
(A) cutting out time window data of a set time length from the sensor data; and
(B) extracting from the time window data peak information identifying a peak included in the time window data;
(C) identifying one peak and the immediately preceding peak based on the peak information, and cutting out a portion existing between the two specified peaks in the sensor data as inter-peak data;
(D) From the peak-to-peak data, as a feature quantity serving as an index for evaluating the behavior of the measurement target, a degree of float indicating a degree of occurrence of a state in which a vertical acceleration applied to the measurement target is equal to or less than a set value is calculated. Steps,
(E) determining the behavior of the measurement object based on the feature amount; and
As the feature amount other than (f) the float level, possess the acceleration variance of the time each window data to calculate the distance between the two peaks the specified steps, a,
In the step (e), based on at least one of the floating degree, the acceleration dispersion value, and the interval between the two specified peaks, the action of the measurement target is assumed in advance and stops. , running, walking, climbing stairs, and down the stairs, two or more any in either the appropriate actions selected from, or for one of these behaviors, or corresponds thereto, determining, that An action determination method characterized by In the step (d), the portion of the peak-to-peak data that is equal to or less than the set value is integrated over time, and the obtained integrated value is defined as the float.
The behavior determination method according to claim 15 , wherein in the step (e), the behavior of the measurement target is determined based on a threshold value of the integral value set for each supposed behavior of the measurement target. The behavior determination method according to claim 15 or 16 , wherein, in the step (d), the set value is set to a value of gravitational acceleration. In the step (e), based on the acceleration dispersion value, it is determined whether or not the measurement object is stopped. When it is determined that the measurement object is not stopped, based on the interval between the two specified peaks. When determining whether the measurement object is running or walking, and determining that the measurement object is walking, it is determined whether the measurement object is going up the stairs or going down the stairs based on the floating degree. The behavior determination method according to claim 15 . In the step (e), based on the acceleration dispersion value, a first evaluation value indicating the possibility that the measurement object is stopped is calculated, and based on the interval between the two specified peaks, A second evaluation value indicating the possibility that the measurement target is running or walking is calculated, and the measurement target may be climbing up or down the stairs based on the floating degree The behavior determination method according to claim 15 , wherein a third evaluation value indicating sex is calculated. (G) further comprising the step of normalizing the float using a feature quantity other than the float;
In said step of (e), instead of the floating degree, using said values calculated by normalization by step (g), determines the behavior of the measurement object, either of claims 1 5-19 The behavior determination method according to Crab. In said step of (g), the floating degree, normalized with the distance between the two peaks the identified behavior determining method according to claim 2 0. (H) The method further includes a step of storing the feature amount calculated in the step (d) as history information for a set period.
In said step of (e), using the history information, determining the standard values of the feature amount, using the standard values to determine the behavior of the measurement object, either claim 1 5-2 1 The behavior determination method according to. In said step of (e), the calculated average value of the feature amount included in the history information, using the calculated the average value as the standard value, behavior determining method according to claim 2 2. In the step (e), an average value is calculated for the feature values excluding the latest feature value among the feature values included in the history information, and a difference between the latest feature value and the average value is calculated. sought, if the value of the obtained difference is below the threshold, it stops the determination of behavior of the measurement object, behavior determining method according to claim 2 2. (I) outputting the result of the determination in the step (e) to the outside;
(J) receiving an input of information specifying whether or not the result of the determination is correct output in the step (i);
(K) The information received in the step (j) is compared with the determination result output in the step (e), and the determination is performed in the step (e) from the comparison result. Updating the rules at the time of,
In the step (e) newly executed after the execution of the steps (i) to (k), the action to be measured is determined using the rules updated in the step (k). behavior determining method according to any one of claims 1 to 5, 21 to 24. (L) The feature amount, the determination result based on the feature amount in the step (e), and the time when the determination in the step (e) is performed are used as history information during a set period. Remember, step,
(M) outputting the determination result and the time for the set period stored in step (l) to the outside;
(N) receiving an input of information specifying whether the determination result and the time are correct or not, which is output in the step (m);
(O) The information received in the step (n) is compared with the determination result and the time output in the step (m). From the comparison result, the information in the step (e) is compared. Updating a rule for making a determination, and
In the step (e) newly executed after the execution of the steps (l) to (o), the action to be measured is determined using the rule updated in the step (o). The behavior determination method according to any one of claims 15 to 19 . A program for executing a determination of an action of a measurement target based on time-series sensor data output from an acceleration sensor by a computer,
In the computer,
(A) cutting out time window data of a set time length from the sensor data; and
(B) extracting from the time window data peak information identifying a peak included in the time window data;
(C) identifying one peak and the immediately preceding peak based on the peak information, and cutting out a portion existing between the two specified peaks in the sensor data as inter-peak data;
(D) From the peak-to-peak data, as a feature quantity serving as an index for evaluating the behavior of the measurement target, a degree of float indicating a degree of occurrence of a state in which a vertical acceleration applied to the measurement target is equal to or less than a set value is calculated. Steps,
(E) determining the behavior of the measurement object based on the feature amount; and
(F) calculating an acceleration dispersion value for each time window data and an interval between the specified two peaks as a feature amount other than the floating degree ; and
In the step (e), based on at least one of the floating degree, the acceleration dispersion value, and the interval between the two specified peaks, the action of the measurement target is assumed in advance and stops. , running, walking, climbing stairs, and down the stairs, two or more any in either the appropriate actions selected from, or for one of these behaviors, or corresponds to it, you determine, A program characterized by that. In the step (d), the portion of the peak-to-peak data that is equal to or less than the set value is integrated over time, and the obtained integrated value is defined as the float.
28. The program according to claim 27 , wherein in the step (e), the behavior of the measurement target is determined based on a threshold value of the integral value set for each supposed behavior of the measurement target. The program according to claim 27 or 28 , wherein in the step (d), the set value is set to a value of gravitational acceleration. In the step (e), based on the acceleration dispersion value, it is determined whether or not the measurement object is stopped. When it is determined that the measurement object is not stopped, based on the interval between the two specified peaks. When determining whether the measurement object is running or walking, and determining that the measurement object is walking, it is determined whether the measurement object is going up the stairs or going down the stairs based on the floating degree. The program according to claim 27 . In the step (e), based on the acceleration dispersion value, a first evaluation value indicating the possibility that the measurement object is stopped is calculated, and based on the interval between the two specified peaks, A second evaluation value indicating the possibility that the measurement target is running or walking is calculated, and the measurement target may be climbing up or down the stairs based on the floating degree The program according to claim 27 , wherein a third evaluation value indicating sex is calculated. In addition to the computer,
(G) performing the step of normalizing the float using a feature quantity other than the float;
In said step of (e), instead of the floating degree, by using the value calculated by the normalization by said step of (g), it determines the behavior of the measurement object, either Claim 27-3 1 The program described in In said step of (g), the floating degree, normalized with the distance between the two peaks the identified program according to claim 3 2. In addition to the computer,
(H) storing the feature amount calculated in the step (d) as history information for a set period,
In said step of (e), using the history information, a standard value of said characteristic quantity calculated, by using the standard value to determine the behavior of the measurement target, in any one of claims 27 to 3 3 The program described. 35. The program according to claim 34 , wherein in the step (e), an average value of the feature amounts included in the history information is calculated, and the calculated average value is used as the standard value. In the step (e), an average value is calculated for the feature values excluding the latest feature value among the feature values included in the history information, and a difference between the latest feature value and the average value is calculated. The program according to claim 34 , wherein the determination of the behavior to be measured is stopped when the obtained difference value is equal to or less than a threshold value. In addition to the computer,
(I) outputting the result of the determination in the step (e) to the outside;
(J) receiving an input of information specifying whether or not the result of the determination is correct output in the step (i);
(K) The information received in the step (j) is compared with the determination result output in the step (e), and the determination is performed in the step (e) from the comparison result. Updating the rules when the process is executed, and
In the step (e) newly executed after the execution of the steps (i) to (k), the action to be measured is determined using the rules updated in the step (k). The program according to any one of claims 27 to 36 . In addition to the computer,
(L) The feature amount, the determination result based on the feature amount in the step (e), and the time when the determination in the step (e) is performed are used as history information during a set period. Remember, step,
(M) outputting the determination result and the time for the set period stored in step (l) to the outside;
(N) receiving an input of information specifying whether the determination result and the time are correct or not, which is output in the step (m);
(O) The information received in the step (n) is compared with the determination result and the time output in the step (m). From the comparison result, the information in the step (e) is compared. Update the rules for making a decision, execute the step,
In the step (e) newly executed after the execution of the steps (l) to (o), the action to be measured is determined using the rule updated in the step (o). The program according to any one of claims 27 to 31 .

Images