JP7718913B2 - Gait analysis system, gait analysis method, and gait analysis program - Google Patents

Description

The present invention relates to a gait analysis system, a gait analysis method, and a gait analysis program.

Conventionally, methods have been known to acquire information on people's movements in public places, such as their direction, speed, and length of stay, for purposes such as optimizing product layout and predicting busy periods. For example, systems such as Omron's HVC-C2W recognize people based on image information captured by a camera and acquire information on each person's movement.

For example, Patent Document 1 discloses a behavior recognition device that improves the accuracy of recognizing the behavior of an object based on the detection results of multiple detection means, without user intervention. In this behavior recognition device, a motion detection unit analyzes images captured by a camera (first sensor) to detect the movement of a person (object), and the motion detection unit analyzes the output of a sensor tag (second sensor) to detect the movement of the person. A reliability calculation unit then assigns a reliability to the detection result of the movement of the person based on the output of the sensor tag. The reliability calculation unit also assigns a reliability to the detection result of the movement of the person based on the image. Furthermore, the behavior recognition unit recognizes the behavior of the person without user intervention, based on the detection result of the movement of the person detected by the motion detection unit and the reliability of the detection result. This enables the behavior of the object to be recognized with high accuracy.

Japanese Patent Application Laid-Open No. 2019-159726

However, the behavior recognition device described in Patent Document 1 has the problem of portrait rights issues arising when the captured image includes the face of a pedestrian. Furthermore, because complex processing such as calculating reliability is required, there is also the problem of long processing times, particularly when recognizing the behavior of a large number of people.

The present invention was developed in consideration of these issues, and its primary objective is to provide a gait analysis system that can analyze information about the gait of an unspecified number of pedestrians with high accuracy and in a short time, without causing issues with portrait rights. It also aims to provide an analysis method and gait analysis program that can analyze information about the gait of an unspecified number of pedestrians with high accuracy and in a short time.

The present invention employs the following measures to achieve at least one of the above-mentioned objectives.

The gait analysis system of the present invention comprises:
pressure information input means for inputting pressure information output from a pressure information measurement means for measuring pressure information including pressure values from the soles of the feet of a walker, and information regarding measurement times and measurement locations of the pressure values;
a pressure information classification means for classifying the pressure information input to the pressure information input means into pressure information for each arbitrary foot shape;
a foot shape classifying means for classifying the arbitrary foot shapes classified by the pressure information classifying means into foot shapes for arbitrary persons;
characterized in that it comprises
It is something.

This gait analysis system uses a pressure information classification means to classify pressure information input into a pressure information input means that receives pressure information output from a pressure information measurement means that measures pressure information including pressure values from the soles of the walker's feet, the measurement time and measurement location of those pressure values, into pressure information for each arbitrary foot shape. The arbitrary foot shapes classified by the pressure information classification means are then classified into foot shapes for each arbitrary person using a foot shape classification means. This makes it possible to obtain information that can be used to analyze information about a walker's gait with high accuracy and in a short time, using only the pressure information from the soles of the walker's feet, without causing any issues with the walker's portrait rights.

The gait analysis system of the present invention may also be characterized by including a gait information analysis means that analyzes gait information for each person based on the foot shapes classified for that person by the foot shape classification means. This makes it possible to analyze information about a pedestrian's gait with high accuracy and in a short time using only pressure information derived from the soles of the pedestrian's feet, without causing any issues with the pedestrian's portrait rights.

In a gait analysis system of the present invention employing this aspect, the gait information analysis means may be characterized by calculating at least one of the stride length, stride width, walking speed, or walking cycle for each person from the foot shape classified for that person by the foot shape classification means. This makes it possible to analyze information about a pedestrian's gait with high accuracy and in a short time, and calculate at least one of the stride length, stride width, walking speed, or walking cycle for each person, using only pressure information derived from the soles of the pedestrian's feet, without causing any issues with the pedestrian's portrait rights.

In the gait analysis system of the present invention, the pressure information classification means may classify the pressure information into pressure information for any given foot shape by clustering pressure information derived from the same foot shape using a clustering algorithm that utilizes the K-means method. This makes it possible to analyze information about a pedestrian's gait with high accuracy and in a short time using only pressure information derived from the soles of the pedestrian's feet, without causing any issues with the pedestrian's portrait rights.

In the gait analysis system of the present invention, the foot shape classification means may classify the foot shapes of any individual based on a pressure distribution transition vector. This allows for highly accurate and rapid analysis of information about a pedestrian's gait, without violating the pedestrian's portrait rights, using only pressure information from the soles of the feet. Note that the "pressure distribution transition vector" refers to a vector (a vector in three-dimensional spatial coordinates consisting of X coordinates, Y coordinates, and time t) that is calculated by calculating the center of pressure or area center in the frames immediately before the clusters appear and disappear from the clustered pressure value point cloud, calculating the point on the coordinates immediately before the clusters disappear, and calculating the end point from the point immediately before the clusters disappear. This pressure distribution transition vector is a three-dimensional vectorization using the spread of reaction points from the clusters' appearance to their disappearance, the trajectory of the center of foot pressure, and other data.

In the gait analysis system of the present invention, the foot shape classification means may classify the foot shapes of any person based on the reference estimated existence range. This makes it possible to analyze information about a pedestrian's gait with high accuracy and in a short time, using only pressure information from the soles of the pedestrian's feet, without causing any issues with the pedestrian's portrait rights. Note that here, the "reference estimated existence range" refers to the range in which the next foot shape is likely to be located.

The gait analysis method of the present invention comprises:
a pressure information input step of inputting pressure information output from a pressure information measurement means that measures pressure information including a pressure value from the sole of a walker, and information regarding a measurement time and a measurement location of the pressure value;
a pressure information classification step of classifying the pressure information inputted in the pressure information input step into pressure information for each arbitrary foot shape;
a foot shape classification step of classifying the arbitrary foot shapes classified in the pressure information classification step into foot shapes for arbitrary persons;
characterized in that it comprises
It is something.

This gait analysis method involves inputting pressure information output from a pressure information measurement means that measures pressure information including pressure values from the soles of the walker's feet, and information about the measurement time and measurement location of the pressure values, in a pressure information input step, classifying the input pressure information into pressure information for any given foot shape in a pressure information classification step, and then classifying the any foot shapes classified in the pressure information classification step into foot shapes for any given person in a foot shape classification step.By doing so, it is possible to obtain information that can be used to analyze information about a walker's gait with high accuracy and in a short time, from only the pressure information from the soles of the walker's feet, without causing any issues with the walker's portrait rights.

The program of the present invention is a program for causing one or more computers to execute each step of the gait analysis method. This program may be recorded on a computer-readable storage medium (e.g., a hard disk, ROM, CD, DVD, flash memory, etc.), or may be transmitted from one computer to another via a transmission medium (a communication network such as the Internet or a wired/wireless LAN), or may be transmitted in any other form. Furthermore, even if the program is executed on a device that executes each step of the analysis method, the device on which the program is executed may be different from the device on which the processing is performed. In either case, by executing this program on a single computer or by dividing the execution of each step among multiple computers, the same effects as the analysis method described above can be achieved.

FIG. 1 is a block diagram showing an example of electrical connections in a gait analysis system 20 of the present invention. FIG. 2 is an explanatory diagram showing an example of the pressure information obtaining means 10 and pressure information. FIG. 3 is a flowchart showing an example of a gait analysis processing routine of the present invention. FIG. 4 is a flowchart showing an example of raw data processing in the gait analysis processing routine. FIG. 5 is a flowchart showing an example of the foot shape classification process in the gait analysis process routine. FIG. 6 is a schematic diagram showing clustered pressure information. FIG. 7 is a flowchart showing an example of a person-by-person classification process in the gait analysis process routine. FIG. 8 is a flowchart showing an example of the gait analysis process in the gait analysis process routine.

Next, a gait analysis system 20 will be described in detail as an example embodiment of the present invention. The embodiments and drawings described below exemplify some of the embodiments of the present invention and are not intended to limit the scope of the present invention. Appropriate modifications can be made without departing from the spirit of the present invention. Corresponding components in each drawing are designated by the same or similar reference numerals. Furthermore, by showing an example of an analysis method using the gait analysis system 20, an example of a gait analysis method and gait analysis program of the present invention will be clarified, as well as an example of how to use the gait analysis system 20.

As shown in Figure 1, a gait analysis system 20, an example of an embodiment of the present invention, comprises: a pressure information input means 22 that inputs pressure information output from a pressure information measurement means 10 that measures pressure information including pressure values derived from the soles of a walker's feet and information regarding the measurement time and measurement location of the pressure values; a control means 30 (corresponding to the pressure information classification means and foot shape classification means of the present invention) that classifies the pressure information input to the pressure information input means 22 into pressure information for each arbitrary foot shape and classifies the classified arbitrary foot shapes into foot shapes for each arbitrary person; and an output means 24 that outputs the foot shape information for each arbitrary person classified by the control means 30, all of which are electrically connected. This gait analysis system 20 receives pressure information derived from the soles of a walker's feet output from the pressure information measurement means 10 via the pressure information input means 22, and the control means 30 classifies this pressure information into pressure information for each arbitrary foot shape and classifies it into foot shapes for each arbitrary person, and outputs it from the output means 24. This makes it possible to obtain information that can be used to analyze information about a walker's gait with high accuracy and in a short time, from only the pressure information derived from the soles of the walker's feet, without causing issues with the walker's portrait rights.

The pressure information measurement means 10 is a known pressure sensor, and may be, for example, a resin sheet equipped with multiple pressure sensors spaced at predetermined intervals. When a pedestrian walks on this resin sheet, the pressure sensors located on the soles of the pedestrian's feet detect the pedestrian's movements and output position information, including information about the position of the detected pressure sensor, time information, including information about the time of detection, and pressure information, including the detected pressure value, to the pressure information input means 22. In this manner, the control means 30 can acquire information about the pedestrian's movements based on this pressure information. Note that, as shown in FIG. 2(A), the multiple pressure sensors are arranged in a grid pattern, and the positions of the pressure sensors are represented by coordinate positions on an X-Y coordinate system. This pressure sensor position information, pressure values detected by the pressure sensors, and information about the time of detection are output to the pressure information input means 22 as pressure information in a table format in which coordinate positions correspond to pressure values for each time period, as shown in FIG. 2(B). Here, FIG. 2(A) is a schematic diagram showing an example of the pressure information measurement means 10, and FIG. 2(B) is an explanatory diagram showing an example of pressure information.

The pressure information input means 22 is a known input/output means that outputs a signal including the pressure information output from the pressure information measurement means 10 to the control means 30.

The output means 24 is a known output means that outputs a signal containing information about the pedestrian's walking output from the control means 30. The output means 24 may also serve as a display means that outputs and displays the signal containing information about the pedestrian's walking on a monitor, printer, etc.

The control means 30 is configured as a microprocessor centered around a CPU 31, and is electrically connected via a bus 35 to a ROM 32 that stores various processing routines, such as a gait analysis processing routine that classifies pressure information by arbitrary foot shape and by arbitrary person, a RAM 33 that temporarily stores pressure information input from the pressure information input means 22, and an interface 34 (hereinafter referred to as "I/F 34") that sends and receives various signals between the pressure information input means 22, output means 24, etc. This control means 30 classifies pressure information input from the pressure information input means 22 by arbitrary person, and outputs information related to the pedestrian's gait from the output means 24.

Next, the operation of classifying pressure information input from the pressure information input means 22 will be described using as an example a gait analysis processing routine executed by the control means 30. This gait analysis processing routine is executed repeatedly at predetermined times (for example, when a start signal is received or when a signal containing pressure information is received). Note that Figure 3 is a flowchart showing an example of a gait analysis processing routine.

When this gait analysis processing routine is executed by the CPU 31, raw data processing (step S100) is executed, which processes the pressure information input to the pressure information input means 22 into data suitable for subsequent processing, and converts it into data in a format suitable for the foot shape classification processing (step S200) described below. In this raw data processing, the information required for the foot shape classification processing (step S200) is processed from the input pressure information (see Figure 2 (B)).

Now, raw data processing will be described in detail using Figure 4. Note that Figure 4 is a flowchart showing an example of raw data processing. When raw data processing is executed by CPU 31, the value 0 is assigned to variable t, which indicates the table number of the pressure information (step S110), and the pressure information contained in the tth table is temporarily stored in RAM 33 (step S120). Next, CPU 31 determines whether the pressure value stored in the pressure information is equal to or greater than a predetermined threshold (step S130), and if it determines that the pressure value is less than the predetermined threshold, assigns the value 0 to the pressure value (step S140). In this way, by setting values less than the predetermined threshold to 0, pressure values measured using pressures not derived from the foot shape can be eliminated, thereby reducing computational effort and time compared to performing subsequent processing based on all pressure values.

Next, if the CPU 31 determines in step S130 that the pressure value is equal to or greater than the threshold value, or after step S140 is executed, it determines whether step S130 has been executed for all pressure values, and if it determines that step S130 has not been executed for all pressure values, it executes step S130 for the other pressure values. By doing this, all pressure values lower than the threshold value can be replaced with a value of zero.

On the other hand, if the CPU 31 determines in step S150 that processing has been completed for all pressure values, it determines whether all pressure values are zero (step S160), and if it determines that all pressure values are zero, it deletes the pressure table. Since such a pressure table clearly does not contain pressure values measured using pressure derived from the foot shape, deleting this pressure table reduces the effort required to perform the processing described below and shortens the calculation time.

Next, if the CPU 31 determines in step S160 that any of the pressure values is not zero, or after step S170 is executed, it adds 1 to variable t (step S180) and determines whether the value of variable t is equal to or greater than the number of tables. If it determines that the value is less than the number of tables, it executes step S130 again. By doing this, for all tables in which pressure values are stored for each time period, only tables for times when pressure values equal to or greater than the threshold value are stored are extracted, and pressure values less than the threshold value are set to 0, thereby organizing the information into the necessary form, reducing the effort required for the processing described below and shortening the calculation time.

As shown in FIG. 3, following raw data processing (step S100), the CPU 31 executes foot shape classification processing (step S200). In this foot shape classification processing, a collection of pressure values is clustered for each pressure value derived from the same foot shape. In this way, pressure values can be clustered as a collection of pressure values derived from foot shapes, and the collection of clustered pressure values can be treated as the pressure value for one walking step.

The foot shape classification process will now be described in detail using Figure 5. Note that Figure 5 is a flowchart showing an example of the foot shape classification process. When the foot shape classification process is executed by the CPU 31, a three-dimensional clustering process is executed (step S210), and the pressure information is clustered. Specifically, clustering is performed using k-means clustering, a non-hierarchical clustering algorithm. At this time, the clustered data is represented in foot shape position and time space by a set of pressure values, as shown in Figure 6. Note that Figure 6 is a schematic diagram showing the clustered pressure information, showing the positional transition on the X-Y coordinate plane and the time transition in the height direction.

Next, the CPU 31 assigns consecutive cluster numbers starting from 1 to each cluster clustered in step S210 (step S220) and assigns the value 1 to variables n and m, which indicate the cluster numbers (step S230). Next, it determines whether the size of the nth cluster is equal to or greater than a pre-stored threshold (step S240). If it is determined that the size is equal to or greater than the threshold, it calculates a pressure distribution transition vector (step S250). Note that the "cluster size" here can refer to, for example, the area of the X-Y coordinate plane when represented in the schematic diagram shown in FIG. 6. Since the area of this X-Y coordinate plane represents the size of the footprint, using the area of the X-Y coordinate plane as the "cluster size" can eliminate clusters resulting from small objects such as canes. Additionally, the height value when represented in the schematic diagram shown in FIG. 6 can also be used as the "cluster size." Since this height value represents the time the footprint remains on the ground, using the height value as the "cluster size" can eliminate clusters resulting from objects that remain on the ground for a short period of time that would not occur during normal walking. Additionally, "pressure distribution transition vector" refers to the vector (a vector in three-dimensional spatial coordinates consisting of X coordinates, Y coordinates, and time t) from the start point to the end point calculated by calculating the pressure center point or area center point in the frames when the cluster of clustered pressure values occurs and immediately before it disappears, with the point on the coordinates on the occurrence side of the cluster being the start point, and the point on the side immediately before it disappears being the end point. This pressure distribution transition vector is a three-dimensional vector created using the spread of reaction points from the occurrence to disappearance of the cluster, the trajectory of the center of foot pressure, etc.

Next, the CPU 31 stores the nth cluster information and the nth pressure distribution transition vector in the mth structure array (step S260), and adds a value of 1 to the variable n and a value of 1 to the variable m (step S270). By doing so, the cluster information and the corresponding pressure distribution transition vector can be stored in association with the structure array.

On the other hand, if the CPU 31 determines in step S240 that the size of the nth cluster is less than the threshold, it adds a value of 1 to the variable n and a value of 1 to the variable m (step S270). In this way, if the size of the cluster is less than the threshold, it is determined that the cluster is not a cluster of pressure values derived from footprints, and there is no need to calculate the pressure distribution transition vector of such a cluster.

Next, the CPU 31 determines whether the value of variable n is equal to or greater than the total number of clusters (step S280), and if it determines that it is less than the total number of clusters, executes step S240 again. By doing so, if the cluster size is equal to or greater than the threshold for all clusters, the pressure distribution transition vector can be calculated and stored in the structure array.

On the other hand, if it is determined in step S280 that the value of variable n is less than the total number of clusters, the CPU 31 sorts the structure array in ascending order of variable t (step S290) and terminates the foot shape classification process. This makes it possible to obtain structure array data over time that stores clusters and their corresponding pressure distribution transition vectors.

As shown in FIG. 3, the CPU 31 executes the foot print classification process (step S200) followed by the person classification process (step S300). In this person classification process, the cluster representing one footprint obtained in the foot print classification process is classified by person. This allows the foot print to be classified by pedestrian, and multiple footprints to be classified by any pedestrian, thereby obtaining information about the gait of each classified pedestrian.

The person-specific classification process will now be described in detail using Figure 7. Figure 7 is a flowchart showing an example of the person-specific classification process. When the person-specific classification process is executed by the CPU 31, the value 1 is assigned to each of variables j and k (step S310), and the jth cluster is registered as a reference cluster (step S320) and temporarily stored in RAM 33. Note that this variable j refers to a variable corresponding to each cluster number classified in the foot shape classification process.

Next, the CPU 31 determines whether the reference cluster temporarily stored in RAM 33 in step S320 is in a stopped state (step S330). If it determines that the reference cluster is in a stopped state, it registers the estimated presence range for the stopped state as the reference estimated presence range (step S340). If it determines that the reference cluster is not in a stopped state, it registers the estimated presence range for the walking state as the reference estimated presence range (step S350). Specifically, it determines whether the value of the time component of the pressure distribution transition vector included in the reference cluster is equal to or greater than a predetermined threshold (e.g., 2000 ms or 1500 ms). If it is equal to or greater than the threshold, it determines that the object is in a stopped state. If it is less than the threshold, it determines that the object is not in a stopped state. Experiments by the inventors have confirmed that in a normal walking state, the value of the time component of the pressure distribution transition vector is around 1000 ms at the longest. Therefore, by setting the threshold greater than 1000 ms, it is possible to determine whether the object is in a walking state or a stopped state. Note that here, "estimated presence range" refers to the range in which the next footprint is likely to be located. By using different estimated presence ranges depending on whether the cluster derived from the current footprint is in a stationary state or a walking state, more accurate and faster processing can be performed when classifying clusters derived from the same person's footprints.

Here, we will explain the "estimated presence range" in more detail. This estimated presence range refers to the range in which the next footstep is likely to be located. It means a different range for the walking state and the stationary state. Specifically, the walking state's estimated presence range refers to a predetermined range centered on a point a predetermined distance away from the center point of the pressure distribution transition vector in the direction of the pressure distribution transition vector. The stationary state's estimated presence range refers to a position of any predetermined shape (e.g., a rectangular shape or a sector shape with its center point in the direction of the center point of the pressure distribution transition vector) centered on the center point of the pressure distribution transition vector. Typically, when stationary, both feet are positioned close together, and when walking, both feet are positioned a stride apart. Therefore, by defining the estimated presence range for the stationary state and the estimated presence range for the walking state as different positions and using different estimated presence ranges for the stationary state and the walking state, the target cluster can be set to an appropriate range when extracting a cluster derived from the footstep of the next step, resulting in higher accuracy and faster processing.

Next, the CPU 31 determines whether the cluster information classified in the foot print classification process is present within the reference estimated presence range (step S360), and if it determines that cluster information is present within the reference estimated presence range, it determines whether the cluster information determined to be present within the reference estimated presence range is presumed to be the same person as the reference cluster (step S370). In this way, by determining whether only the cluster information present within the reference estimated presence range is presumed to be the same person, processing can be performed faster than when making an estimation for all cluster information.

Here, we will explain in detail the method for determining whether a reference cluster is estimated to represent the same person. For example, this estimation method involves comparing the direction of the pressure distribution transition vector contained in the reference cluster with the direction of the pressure distribution transition vector contained in the cluster information to be compared. If one of the pressure distribution transition vectors points leftward and the other points rightward relative to the direction of travel, it may be inferred that the clusters are the same person. Pressures derived from the footprints of the same person always appear in left-right order. Therefore, if the clusters are located within the estimated presence range and the pressure distribution transition vectors point leftward and rightward relative to the direction of travel, it is highly likely that the clusters are derived from the footprints of the same person. Furthermore, if the difference between the pressure values contained in the reference cluster information and the pressure values contained in the cluster information to be compared is below a threshold, it may be inferred that the clusters are the same person. Because pressure values derived from footprints of the same person are similar, a small difference in pressure values is highly likely that the clusters are derived from the footprints of the same person. Furthermore, if the j-1th cluster information is available, it may be inferred that the clusters are the same person if the footprint shapes contained in the j-1th cluster information and the j+1th cluster information are similar. This is because similar foot shapes are likely to be clusters derived from the same person's foot shapes. Here, "similar" may be determined, for example, when the distribution of pressure values in the cluster (e.g., the distribution on the X-Y coordinate plane when represented in the schematic diagram shown in FIG. 6 ) is highly similar. Because the distribution on the X-Y coordinate plane is derived from the contours of the derived footprints, a high similarity in the distribution of pressure values indicates a high probability that the clusters originate from the same person. Additionally, a high similarity in the distribution of high and low pressure values may be determined to be "similar." When walking, the location of high (or low) pressure values often depends on personal habits and habits. Therefore, a high similarity in the distribution of high and low pressure values indicates a high probability that the clusters originate from the same person. Alternatively, after determining which part of the foot (e.g., the heel, the toe, etc.) corresponds to each pressure value, the time-varying trend of the pressure values (e.g., the transition over time of the sum of the pressure value responses constituting the cluster from the time the foot touches the ground to the time the foot leaves the ground) may be used. Furthermore, two or more of these may be combined for determination, or each may be used as a parameter, weighted, and combined for determination.

If the CPU 31 determines in step S370 that the cluster information is presumed to be the same person as the reference cluster, it registers the cluster information as the kth person information and deletes that information from the cluster list (step S380). This allows the CPU 31 to extract cluster information that is presumed to be the same person as the kth reference cluster and to exclude it from subsequent processing, thereby speeding up the process.

After executing step S380, or if the CPU 31 estimates in step S370 that the person is not the same as the reference cluster, it determines whether there is other cluster information within the reference estimated existence range (step S390), and if it determines that there is other cluster information, it executes step S370 again for the other cluster information. By doing this, if there is multiple cluster information within the reference estimated existence range, it can determine for each cluster information whether it is estimated to be the same person as the reference cluster.

On the other hand, if the CPU 31 determines in step S390 that no other cluster information exists, it adds a value of 1 to each of variables j and k (step S400), and determines whether variable j is equal to or greater than constant J, which is the total number of clusters extracted in the foot shape classification process (step S410). If it determines that variable j is less than constant J, it executes step S320 again. On the other hand, if it determines in step S410 that variable k is equal to or greater than constant J, it terminates this routine. In this way, all cluster information classified in the foot shape classification process can be classified by person.

As shown in Figure 3, the CPU 31 executes the gait analysis process (step S500) following the person classification process (step S300). In this gait analysis process, the walking state of each person obtained in the person classification process is analyzed. In this way, information about the gait of each walker can be obtained from the multiple foot shapes classified for each person.

The gait analysis process will now be described in detail using Figure 8. Figure 8 is a flowchart showing an example of the gait analysis process. When the gait analysis process is executed by the CPU 31, the value 1 is assigned to the variable h (step S510), and the cluster information of the hth person classified in the person-by-person classification process is temporarily stored in RAM 33 (step S520). Next, the stride length and step width are calculated from the cluster information of the hth person temporarily stored in RAM 33 (step S530). The cluster information of the hth person includes cluster information corresponding to the hth person classified in the person-by-person classification process, out of the foot shapes classified in the foot shape classification process, and therefore the stride length and step width can be calculated based on the respective cluster information.

Next, the CPU 31 calculates the walking speed and walking period (step S540). The cluster information for the hth person includes cluster information corresponding to the hth person classified in the person-by-person classification process among the footprints classified in the foot-shape classification process, so the walking speed and walking period can be calculated based on the stride length, stride width calculated in step S530, and the measurement interval derived from the cluster information.

Next, the CPU 31 calculates the progression of the trajectory of the center of pressure (step S550). The cluster information for the hth person includes cluster information corresponding to the hth person classified by the person classification process among the footprints classified by the foot shape classification process, so the trajectory of the center of pressure can be calculated.

Next, the CPU 31 calculates the pressure transition for each step and each body part (step S560). The cluster information for the hth person includes the cluster information corresponding to the hth person classified by the person classification process among the footprints classified by the foot shape classification process, so it is possible to calculate the pressure transition for each step and each body part.

Next, the CPU 31 adds the value 1 to the variable h (step S570), and determines whether the variable h is equal to or greater than the person number calculated in the person-by-person classification process (final value of the variable k) (step S580). If it determines that the variable h is less than the person number, it executes step S530 again; if it determines that the variable h is equal to or greater than the person number, it terminates this routine. In this way, it is possible to calculate walking information for cluster information corresponding to all people classified in the person-by-person classification process.

According to the embodiment of the gait analysis system 20 described above, the pressure information input to the pressure information input means 22, which receives pressure information output from the pressure information measurement means 10 that measures pressure information including information about the pressure value, measurement time, and measurement location from the soles of the walker's feet, is classified by a pressure information classification process (step S200) that classifies the pressure information into pressure information for each arbitrary foot shape, and the arbitrary foot shapes classified by the pressure information classification process are then classified into foot shapes for each arbitrary person by a foot shape classification process (step S300).This makes it possible to obtain information about the walker's gait from only the pressure information from the soles of the walker's feet, without causing any issues with the walker's portrait rights.

Furthermore, in step 500, by using a gait analysis process to infer the behavioral characteristics of a person, including their stride length, stride width, walking speed, and walking cycle, from the foot shapes classified for each person by the foot shape classification process, information about the pedestrian's gait can be obtained from only the pressure information derived from the soles of the pedestrian's feet, without causing any issues with the pedestrian's portrait rights, and by using this information about gait, the behavioral characteristics of a person can be estimated.

Furthermore, in step S200, a clustering algorithm using the K-means method is used to cluster pressure information derived from the same foot shape, thereby classifying the pressure information into pressure information for any given foot shape. This makes it possible to obtain information about a pedestrian's gait with high accuracy and in a short time from only the pressure information derived from the soles of the pedestrian's feet, without causing any issues with the pedestrian's portrait rights.

Furthermore, in step S200, by classifying the foot shape of any person based on the pressure distribution transition vector, information about the pedestrian's gait can be obtained with high accuracy and in a short time from only the pressure information derived from the soles of the pedestrian's feet, without causing any issues with the pedestrian's portrait rights.

Then, in step S200, by classifying the foot shape of each person based on the reference estimated existence range, information about the pedestrian's gait can be obtained with high accuracy and in a short time from only the pressure information derived from the soles of the pedestrian's feet, without causing any issues with the pedestrian's portrait rights.

It goes without saying that the present invention is in no way limited to the above-described embodiments, and can be implemented in various forms as long as they fall within the technical scope of the present invention.

For example, in the embodiment described above, step S100 is executed before step S200 is executed, but step S100 does not have to be executed. Even in this case, the same effects as those of the embodiment described above can be obtained.

In the above-described embodiment, step S500 is executed after step S300, but step S500 does not have to be performed. By analyzing the data output in step S300 using a known information processing device and information processing method, information regarding pedestrian characteristics can be calculated.

As shown in the above-mentioned embodiment, it can be used in the information processing field, particularly as a gait analysis system.

10... Pressure information measurement means, 20... Gait analysis system, 22... Pressure information input means, 24... Output means, 30... Control means, 31... CPU, 32... ROM, 33... RAM, 34... Interface, 35... Bus.

Claims (6)

pressure information input means for inputting pressure information output from a pressure information measurement means for measuring pressure information including pressure values from the soles of the feet of a walker, and information regarding measurement times and measurement locations of the pressure values;
a pressure information classification means for classifying the pressure information input to the pressure information input means into pressure information for each arbitrary foot shape;
a foot shape classifying means for classifying the arbitrary foot shapes classified by the pressure information classifying means into arbitrary foot shapes for each person based on a pressure distribution transition vector ;
characterized in that it comprises
Gait analysis system. The gait analysis system according to claim 1 ,
a gait information analysis means for analyzing gait information of an arbitrary person from the foot shapes classified for each arbitrary person by the foot shape classification means;
characterized in that it comprises
Gait analysis system. the walking information analysis means calculates at least one of a stride length, a step width, a walking speed, and a walking cycle for each of the arbitrary persons from the foot shapes classified for each of the arbitrary persons by the foot shape classification means,
The directional analysis system of claim 2 . The pressure information classification means classifies the pressure information into pressure information for each arbitrary foot shape by clustering the pressure information derived from the same foot shape using a clustering algorithm that utilizes the K-means method.
The gait analysis system according to any one of claims 1 to 3. a pressure information input step of inputting pressure information output from a pressure information measurement means that measures pressure information including a pressure value from the sole of a walker's foot, and information regarding a measurement time and a measurement location of the pressure value;
a pressure information classification step of classifying the pressure information inputted in the pressure information input step into pressure information for each arbitrary foot shape;
a foot shape classification step of classifying the arbitrary foot shapes classified in the pressure information classification step into an arbitrary person's foot shape based on a pressure distribution transition vector;
characterized in that it comprises
Gait analysis methods. A program for executing the gait analysis method according to claim 5 on one or more computers.
Rum.