Analysis method, analysis device, analysis system, and analysis program

Abstract

To improve the efficiency of analyzing data acquired from a sensor fixed to a ball.SOLUTION: An analysis system 1000 includes at least one sensor 20 and an analysis device 10. The at least one sensor 20 is fixed to a ball 2 and measures motion data that includes at least one physical quantity that characterizes at least one of translational movement and rotational movement of the ball 2 following a specific motion of a subject 5 using the ball 2. The analysis device 10 determines which at least one label corresponds to the motion data through classification calculation used for at least one identification data given with at least one label. At least one label includes an identifier of a person who has performed the specific motion corresponding to the identification data given with the label.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present disclosure relates to techniques for analyzing data obtained from sensors affixed to a ball. [Background technology]

[0002] Conventionally, configurations for analyzing data acquired from sensors fixed to a ball are known. For example, International Publication No. 2017 / 131133 (Patent Document 1) discloses a system that determines the ball speed, number of rotations, and type of ball based on sensor data acquired from a ball incorporating an acceleration sensor, a gyro sensor, and a magnetic sensor from the start to the end of the pitch. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] International Publication No. 2017 / 131133 Summary of the Invention [Problem to be solved by the invention]

[0004] When a coach instructs a pitcher based on the analysis results of pitching data acquired from a sensor fixed to a ball, the coach needs to identify which pitcher the pitching data belongs to. In this regard, in actual baseball training, one pitcher may use multiple balls, or multiple pitchers may use one ball. In such cases, in order to identify which pitcher the pitching data acquired from the ball belongs to, it is necessary to integrate data identifying the pitcher for each pitch into the total pitching data. Such data integration may reduce the efficiency of the analysis of pitching data. However, the system disclosed in Patent Document 1 does not take into consideration such problems that may occur in actual training.

[0005] The present disclosure has been made to solve the above-mentioned problems, and its purpose is to improve the efficiency of analysis of data obtained from a sensor fixed to a ball. [Means for solving the problem]

[0006] A method according to one aspect of the present disclosure includes the steps of acquiring motion data including at least one physical quantity characterizing at least one of translational motion and rotational motion of the ball associated with a specific motion of a subject using the ball from at least one sensor fixed to the ball, and determining which of the at least one label the motion data corresponds to by a classification calculation using at least one identification data each having at least one label attached thereto, each of the at least one label including information capable of identifying a person who performed the specific motion corresponding to the labeled identification data.

[0007] Preferably, the classification calculation may identify a correct answer candidate data having a maximum similarity between the at least one identification data and the motion feature data extracted from the motion data.

[0008] Preferably, the step of determining which of at least one label the action data corresponds to may determine that the action data corresponds to the label assigned to the correct candidate data when the similarity between the correct candidate data and the action feature data is greater than a discrimination threshold.

[0009] Preferably, the similarity may be calculated by DTW (Dynamic Time Warping).

[0010] Preferably, the classification calculation may classify the motion data into one of at least one class corresponding to at least one label using a classification model that has been trained by machine learning using at least one identification data as training data.

[0011] Preferably, the motion characteristic data may be generated from motion data from a specific time when the subject releases the ball to a time preceding the specific time by a reference time interval.

[0012] Preferably, the at least one sensor may include at least one of an acceleration sensor, a gyro sensor, and a geomagnetic sensor.

[0013] An analysis device according to another aspect of the present disclosure includes a processing unit and a storage unit. The processing unit acquires motion data including at least one physical quantity characterizing at least one of translational motion and rotational motion of the ball associated with a specific motion of a subject using the ball from at least one sensor fixed to the ball. The motion data is stored in the storage unit. The processing unit determines which of the at least one label the motion data corresponds to by a classification calculation used for at least one identification data each having at least one label attached thereto. Each of the at least one label includes information capable of identifying a person who performed the specific motion corresponding to the identification data attached thereto.

[0014] An analysis device according to another aspect of the present disclosure includes at least one sensor and a processing unit. The at least one sensor measures motion data including at least one physical quantity characterizing at least one of translational motion and rotational motion of a ball associated with a specific motion of a subject using the ball. The processing unit determines which of the at least one label the motion data corresponds to by a classification calculation used for at least one identification data each having at least one label attached thereto. Each of the at least one label includes information capable of identifying a person who performed the specific motion corresponding to the labeled identification data.

[0015] An analysis system according to another aspect of the present disclosure includes at least one sensor and an analysis device. The at least one sensor is fixed to a ball and measures motion data including at least one physical quantity that characterizes at least one of translational motion and rotational motion of the ball associated with a specific motion of a subject using the ball. The analysis device determines which of the at least one label the motion data corresponds to by a classification calculation used for at least one identification data each having at least one label attached thereto. Each of the at least one label includes information that can identify a person who performed the specific motion corresponding to the labeled identification data.

[0016] An analysis program according to another aspect of the present disclosure is executed by at least one processor to cause the at least one processor to perform a plurality of operations. The plurality of operations includes a step of acquiring, from at least one sensor fixed to the ball, operation data including at least one physical quantity characterizing at least one of translational motion and rotational motion of the ball associated with a specific operation of a subject using the ball, and a step of determining which of the at least one label the operation data corresponds to by a classification calculation used for at least one identification data each having at least one label attached thereto. Each of the at least one label includes information capable of identifying a person who performed the specific operation corresponding to the identification data attached thereto. Effect of the Invention

[0017] According to the present disclosure, it is possible to efficiently analyze data obtained from a sensor fixed to a ball. [Brief description of the drawings]

[0018] [Figure 1] FIG. 1 is a diagram for explaining an example of an overall configuration of an analysis system according to a first embodiment. [Diagram 2] FIG. 2 is a diagram for explaining an example of a schematic configuration of the ball in FIG. [Diagram 3] 2 is a block diagram showing an example of a hardware configuration of the sensor device in FIG. 1. [Figure 4] 2 is a block diagram showing an example of a hardware configuration of the information processing terminal of FIG. 1. [Diagram 5] 5 is a diagram showing some of the plurality of pitch identification data included in the pitch identification data set of FIG. 4 when the pitch characteristic data of FIG. 1 is resultant acceleration. FIG. [Figure 6] 5 is a diagram showing a portion of a plurality of pitch identification data included in the pitch identification data set of FIG. 4 when the pitch characteristic data of FIG. 1 is a resultant angular velocity. FIG. [Figure 7] 5 is a flowchart showing an example of the flow of a classification process performed by a processor that executes the analysis program of FIG. [Figure 8] FIG. 11 is a block diagram showing an example of a hardware configuration of a sensor device according to a second embodiment. [Figure 9] FIG. 11 is a block diagram showing an example of a hardware configuration of an information processing terminal according to a second embodiment. [Figure 10] 9 is a flowchart showing an example of the flow of a classification process performed by a processor that executes the analysis program of FIG. 8. [Figure 11] FIG. 11 is a block diagram showing an example of a hardware configuration of an information processing terminal according to a third embodiment. [Figure 12] 12 is a flowchart showing an example of the flow of a classification process performed by a processor that executes the analysis program of FIG. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0019] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. In the following description, the same components are denoted by the same reference numerals. Their names and functions are also the same. Therefore, detailed description thereof will not be repeated.

[0020] [Embodiment 1] FIG. 1 is a diagram for explaining an example of the overall configuration of an analysis system 1000 according to the first embodiment. FIG. 1 shows a pitching motion from when a subject 5, who is a pitcher, takes a swing to throw a ball 2 toward a catcher 6, and a trajectory Tr of the ball 2 from when the ball 2 is released by the subject 5 until it is caught by the catcher 6. The analysis system 1000 is a system for analyzing pitch data (flight ball data) of the subject 5 and displaying the analysis results. In FIG. 1, an absolute coordinate system is formed by an X-axis, a Y-axis, and a Z-axis that are orthogonal to each other. The negative direction of the Z-axis of the absolute coordinate system corresponds to the direction of gravity.

[0021] As shown in FIG. 1, the analysis system 1000 includes an information processing terminal 10 (analysis device) and a baseball ball 2. The ball 2 includes a built-in sensor device 20. The coordinate system of the sensor device 20 (sensor coordinate system or local coordinate system) has three mutually orthogonal axes: x-axis, y-axis, and z-axis. The sensor device 20 measures physical quantities (e.g., acceleration, angular velocity, or magnetic field (magnetic flux density)) of the three axes of the sensor coordinate system that characterize at least one of the translational motion and rotational motion of the ball 2 at each sampling time (e.g., 1 ms), and outputs time series data (time chart) of the physical quantities. The time series data of the physical quantities measured by the sensor device 20 during the pitching motion of the subject 5 and while the ball is flying along the trajectory Tr corresponds to the pitching data of the subject 5. The sensor device 20 transmits the pitching data to the information processing terminal 10, for example, at the timing when the ball 2 is caught by the catcher 6.

[0022] The information processing terminal 10 is placed near the catcher 6 and communicates with the sensor device 20 by a wireless communication method. For example, the information processing terminal 10 establishes wireless communication with the sensor device 20 at the timing when the ball 2 is caught by the catcher 6, and receives pitch data from the sensor device 20. For example, BLE (Bluetooth (registered trademark) Low Energy) is adopted as the wireless communication method. However, the information processing terminal 10 may adopt other wireless communication methods such as Bluetooth (registered trademark) and wireless LAN (Local Area Network). The information processing terminal 10 receives the pitch data from the sensor device 20 and displays an analysis result of the pitch data. The information processing terminal 10 includes, for example, a smartphone, a laptop PC (Personal Computer), a tablet terminal, or a desktop PC.

[0023] FIG. 2 is a diagram for explaining an example of a schematic configuration of the ball 2 in FIG. 1. FIG. 2(a) shows the appearance of the ball 2. FIG. 2(b) shows a schematic configuration inside the ball 2. As shown in FIG. 2(a), the appearance of the ball 2 is the same as that of a typical hard ball. The ball 2 has a leather outer skin and is configured so that the stitching is visible. As shown in FIG. 2(b), the ball 2 has a built-in sensor device 20 for detecting the behavior of the ball 2 in its center. The sensor device 20 is fixed by a polycarbonate capsule 62 and a silicone gel 64, and has excellent impact resistance.

[0024] Fig. 3 is a block diagram showing an example of a hardware configuration of the sensor device 20 in Fig. 1. As shown in Fig. 3, the sensor device 20 includes, as main components, a processor 202 (processing unit), a memory 204 (storage unit), a measurement sensor 205, a communication interface (I / F) 210, and a storage battery 212. The storage battery 212 supplies power to various components of the sensor device 20.

[0025] The processor 202 includes, for example, a CPU (Central Processing Unit) or a GPU (Graphics Processing Unit). The processor 202 executes various processes corresponding to the functions of the sensor device 20, and communicates with the information processing terminal 10 via the communication interface 210. The memory 204 is realized by a RAM (Random Access Memory), a ROM (Read-Only Memory), a flash memory, or the like. The memory 204 stores programs executed by the processor 202, data executed and used by the processor 202, processing results of the processor 202, and the like.

[0026] The memory 204 stores a measurement program Pgm for controlling the measurement sensor 205 to measure physical quantities. The processor 202 executing the measurement program Pgm controls the measurement sensor 205 to measure pitch data Dm. The processor 202 transmits the pitch data Dm to the information processing terminal 10 via the communication interface 210. The measurement sensor 205 includes, for example, an acceleration sensor, a gyro sensor, or a geomagnetic sensor. For example, an MR (Magnet resistive) element, an MI (Magnet impedance) element, or a Hall element is used as the geomagnetic sensor.

[0027] When a coach instructs a pitcher based on the analysis results of pitching data acquired from the sensor device 20 built into the ball 2, the coach needs to identify which pitcher the pitching data belongs to. In this regard, in actual baseball training, one pitcher may use multiple balls 2, or multiple pitchers may use one ball 2. In such cases, in order to identify which pitcher the pitching data acquired from the ball 2 belongs to, it is necessary to integrate the data identifying the pitcher for each pitch into the total pitching data. Such data integration may reduce the efficiency of the analysis of the pitching data.

[0028] Therefore, in the analysis system according to the embodiment, a pitch identification data set is prepared, which includes pitch identification data to which a label including a pitcher's name is attached, prior to analysis of pitch data. A classification process based on the pitch identification data set determines whether the pitch data to be analyzed corresponds to any of a plurality of labels included in the pitch identification data set. According to the analysis system according to the embodiment, the pitcher's name corresponding to the pitch data to be analyzed is automatically identified, thereby making it possible to efficiently analyze pitch data.

[0029] The analysis system 1000 uses rule-based similarity calculation in classification processing based on the pitch identification data set. The more similar two pieces of data are, the greater the similarity between them. Therefore, the similarity calculation identifies pitch identification data that has the greatest similarity to pitch feature data generated from the pitch data to be analyzed, among at least one pitch identification data set included in the pitch identification data set. The similarity includes, for example, DTW (Dynamic Time Warping) distance, Manhattan distance (L1 norm or L1 distance), Euclidean distance (L2 norm, L2 distance, or Euclidean distance), Lp norm (generalization of L1 norm and L2 norm), max distance (distance of only the most different dimension), Hamming distance (number of different elements in binary numbers or character strings, etc.), edit distance (dictionary distance), Mahalanobis distance, cosine similarity, or correlation coefficient (for example, Pearson's product moment correlation coefficient). When a distance defined between two data is calculated in a similarity calculation, the smaller the distance, the greater the similarity between the two data. In the following, a case where DTW is used as a rule-based similarity calculation will be described.

[0030] Fig. 4 is a block diagram showing an example of a hardware configuration of the information processing terminal 10 in Fig. 1. As shown in Fig. 4, the information processing terminal 10 includes, as main components, a processor 102 (processing unit), a memory 104 (storage unit), a touch panel 106, a button 108, a display 110, a wireless communication unit 112, a communication antenna 113, a memory interface (I / F) 114, a speaker 116, a microphone 118, and a communication interface (I / F) 120. In addition, a recording medium 115 (storage unit) is a storage medium inserted into the memory interface 114 from the outside.

[0031] The processor 102 includes, for example, a CPU or a GPU. The processor 102 reads out and executes a program stored in the memory 104 to control the operation of each unit of the information processing terminal 10.

[0032] The memory 104 is realized by a RAM, a ROM, a flash memory, etc. The memory 104 stores programs executed by the processor 102, data used by the processor 102, processing results of the processor 102, and the like.

[0033] The touch panel 106 is provided on a display 110 having a function as a display unit, and may be of any type such as a resistive film type or a capacitive type. The buttons 108 are arranged on the surface of the information processing terminal 10, and receive instructions from a user and input the instructions to the processor 102.

[0034] The wireless communication unit 112 connects to a mobile communication network via a communication antenna 113 and transmits and receives signals for wireless communication. This enables the information processing terminal 10 to communicate with a predetermined external device via a mobile communication network such as LTE (Long Term Evolution).

[0035] The memory interface (I / F) 114 reads data from the external recording medium 115. The processor 102 reads data stored in the external recording medium 115 via the memory interface 114 and stores the data in the memory 104. The processor 102 reads data from the memory 104 and stores the data in the external recording medium 115 via the memory interface 114. Note that examples of the recording medium 115 include media that store data in a non-volatile manner, such as a secure digital (SD) card, a compact disc (CD), a digital versatile disc (DVD), a Blu-ray (registered trademark) disc (BD), a universal serial bus (USB) memory, a memory card, a flexible disc (FD), and a hard disk. The recording medium 115 may also be provided in a cloud system.

[0036] The speaker 116 outputs sound based on a command from the processor 102. The microphone 118 accepts speech input to the information processing terminal 10.

[0037] The communication interface (I / F) 120 is, for example, a communication interface for transmitting and receiving data between the information processing terminal 10 and the sensor device 20, and is realized by an adapter, a connector, etc. The communication method is, for example, wireless communication by BLE, wireless LAN, etc. The processor 102 receives the pitch data Dm from the sensor device 20 via the communication interface 120.

[0038] The recording medium 115 stores a pitch identification data set Sdp. The pitch identification data set Sdp may be stored in the memory 104. The memory 104 stores a pitch data set Sdm and an analysis program Pga for analyzing the pitch data Dm. The processor 102 executing the analysis program Pga generates pitch characteristic data from the pitch data Dm and calculates the similarity between the pitch characteristic data and each of the multiple pitch identification data included in the pitch identification data set Sdp. The processor 102 executing the analysis program Pga displays the pitcher name and pitch type corresponding to the pitch data Dm on the display 110 based on the result of the similarity calculation, and stores the pitch data Dm associated with these in the memory 104 as an element of the pitch data set. The pitch data set Sdm may be stored in the recording medium 115.

[0039] Referring again to FIG. 1, the pitch data Dm includes time series data of three physical quantities Pqx(t), Pqy(t), and Pqz(t) corresponding to the x-axis, y-axis, and z-axis of the sensor coordinate system at time t, respectively. The pitch characteristic data Df generated from the pitch data Dm is generated from the pitch data Dm for a time interval from time t1 when the subject 5 released the ball 2 to a timing going back a reference time (for example, 100 ms). Pitchers have various pitching forms, and the pitcher's characteristics tend to appear in the pitching form. By extracting data for that time interval (pitching motion data) from the pitch data Dm, it is possible to reduce the amount of data in the pitch characteristic data Df while reflecting the pitcher's characteristics in the pitch characteristic data Df. The pitch characteristic data Df is, for example, time series data of a composite physical quantity Pqc(t) expressed by the following formula (1).

[0040]

number

[0041] The time t1 when the subject 5 releases the ball 2 is identified as the time when any one of the three differential values ​​of the accelerations on the x-axis, y-axis, and z-axis or the three differential values ​​of the angular velocities included in the pitch data Dm exceeds a release threshold. The release threshold can be appropriately determined by simulation or an experiment using an actual machine.

[0042] Alternatively, the time when any of the inexact differential values ​​D of the three magnetic fields of the x-axis, y-axis, and z-axis included in the pitch data Dm becomes 0 may be identified as time t1. The inexact differential value D is calculated using the following formula (2). In formula (2), s represents the Laplace operator, and n represents the differential coefficient.

[0043]

number

[0044] FIG. 5 is a diagram showing a part of a plurality of pitch identification data included in the pitch identification data set Sdp of FIG. 4 when the pitch characteristic data Df of FIG. 1 is a resultant acceleration. As shown in FIG. 5, the time interval of each of the pitch identification data Dp11, Dp12, and Dp13 is 100 ms. The pitch identification data Dp11, Dp12, and Dp13 are respectively assigned labels Lb11, Lb12, and Lb13. The label Lb11 includes "Pitcher A" and "Straight" as the pitcher name and pitch type corresponding to the pitch identification data Dp11. The label Lb12 includes "Pitcher A" and "Curve" as the pitcher name and pitch type corresponding to the pitch identification data Dp12. The label Lb13 includes "Pitcher B" and "Shoot" as the pitcher name and pitch type corresponding to the pitch identification data Dp13. The identifier of the person included in the label attached to the pitch identification data is not limited to a character string of a name, and may be any information that can identify the person who performed the pitching motion corresponding to the pitch identification data. For example, the information may be an identification number, an identification symbol, or an identification image that identifies the person.

[0045] By using DTW as a rule-based similarity calculation, the time interval of the pitch characteristic data Df does not need to be the same as the time interval of the pitch identification data. For example, even if the time interval of the pitch characteristic data Df is 80 ms, it is possible to calculate the similarity (for example, the reciprocal of the DTW distance) between the pitch characteristic data Df and the pitch identification data Dp13 using DTW. Since it is not necessary to align the time interval between the pitch characteristic data Df and the pitch identification data using DTW, it is possible to efficiently collect pitch identification data and calculate the similarity between the pitch characteristic data Df and the pitch identification data without being restricted by the time interval.

[0046] FIG. 6 is a diagram showing a part of a plurality of pitch identification data included in the pitch identification data set Sdp of FIG. 4 when the pitch characteristic data Df of FIG. 1 is a composite angular velocity. As shown in FIG. 5, the time interval of each of the pitch identification data Dp21, Dp22, and Dp23 is 100 ms. The pitch identification data Dp21, Dp22, and Dp23 are respectively attached with labels Lb21, Lb22, and Lb23. The label Lb21 includes "Pitcher C" and "Curve" as the pitcher name and pitch type corresponding to the pitch identification data Dp21. The label Lb22 includes "Pitcher D" and "Straight" as the pitcher name and pitch type corresponding to the pitch identification data Dp22. The label Lb23 includes "Pitcher E" and "Fork" as the pitcher name and pitch type corresponding to the pitch identification data Dp23.

[0047] If the pitching data is time-series data of a magnetic field measured by a geomagnetic sensor, the angular velocity of the three axes of the sensor coordinate system can be calculated from the magnetic field of the three axes of the sensor coordinate system and the time derivative of the magnetic field of the three axes. Therefore, pitching characteristic data of the resultant angular velocity as shown in Figure 6 can be derived from the magnetic field time-series data.

[0048] FIG. 7 is a flowchart showing an example of the flow of the classification process performed by the processor 102 executing the analysis program Pga of FIG. 4. In the following, steps are simply referred to as S. As shown in FIG. 7, in S101, the processor 102 acquires pitch data Dm from the ball 2, and proceeds to S102. In S102, the processor 102 extracts pitching motion data from the pitching data Dm, and proceeds to S103. In S103, the processor 102 generates pitching characteristic data Df from the pitching motion data, and proceeds to S104. In S104, the processor 102 calculates the similarity between each of the multiple pitching identification data included in the pitching identification dataset Sdp and the pitching characteristic data Df, and proceeds to S105. In S105, the processor 102 identifies correct answer candidate data that has the highest similarity to the pitch characteristic data Df from among the multiple pitch identification data included in the pitch identification dataset Sdp, and proceeds to the process in S106.

[0049] In S106, the processor 102 determines whether the similarity between the pitching characteristic data Df and the correct answer candidate data is greater than the discrimination threshold. If the similarity between the pitching characteristic data Df and the correct answer candidate data is greater than the discrimination threshold (YES in S106), the processor 102, in S107, stores the pitching data Dm associated with the pitcher name and pitch type of the label attached to the correct answer candidate data in the memory 104 as an element of the pitching data set Sdm, and proceeds to processing in S108. In S108, the processor 102 displays the pitcher name and pitch type of the label attached to the correct answer candidate data on the display, and ends processing. If the similarity between the pitching characteristic data Df and the correct answer candidate data is equal to or less than the discrimination threshold (NO in S106), the processor 102, in S109, displays on the display that the pitcher corresponding to the pitching characteristic data Df is not registered, and ends processing.

[0050] The discrimination threshold in S106 can be appropriately determined by simulation or experiment with an actual machine. When the pitch characteristic data Df is a time chart of resultant acceleration, the discrimination threshold is, for example, 100. When the pitch characteristic data Df is a time chart of resultant angular velocity, the discrimination threshold is, for example, 8000. By using the discrimination threshold, when the pitcher characteristic data set Sdp does not contain data of a pitcher corresponding to the pitcher characteristic data Df, it is possible to prevent an erroneous determination that a pitcher other than the pitcher is matched to the pitcher characteristic data Df.

[0051] As described above, the analysis method, analysis device, analysis system, and analysis program according to the first embodiment can improve the efficiency of analysis of data acquired from a sensor fixed to a ball.

[0052] [Embodiment 2] In the first embodiment, a configuration has been described in which pitch data is transmitted from the ball to the analysis device, and the analysis device performs classification processing on the pitch data. In the second embodiment, a configuration will be described in which classification processing on the pitch data is performed on the ball from which the pitch data is obtained.

[0053] Fig. 8 is a block diagram showing an example of a hardware configuration of the sensor device 20B according to the embodiment 2. The configuration of the sensor device 20B (analysis device) is such that an analysis program Pga2, and the pitch identification data set Sdp and pitch data set Sdm of Fig. 4 are added to the memory 204 of Fig. 3, and the pitch data Dm is replaced with the classification result Rc. Other than this, the configuration of the sensor device 20B is similar to the configuration of the sensor device 20, and therefore a description of the similar configuration will not be repeated.

[0054] 8, the processor 202 executing the analysis program Pga2 generates pitch characteristic data from pitch data measured by the measurement sensor 205. The processor 202 executing the analysis program Pga2 calculates the similarity between the pitch characteristic data and each of the multiple pitch identification data included in the pitch identification dataset Sdp. The processor 202 executing the analysis program Pga2 transmits a classification result Rc including the pitcher name and pitch type corresponding to the pitch data to the information processing terminal based on the result of the similarity calculation, and stores the pitch data associated with these in the memory 204 as an element of the pitch dataset Sdm.

[0055] Fig. 9 is a block diagram showing an example of a hardware configuration of information processing terminal 10B according to embodiment 2. In the configuration of information processing terminal 10B, analysis program Pga and pitching data set Sdm are removed from memory 104 in Fig. 4, recording medium 115 is removed, and pitching data Dm is replaced with classification result Rc. Other than this, the configuration of information processing terminal 10B is similar to the configuration of information processing terminal 10, so a description of the similar configuration will not be repeated. Processor 102 displays classification result Rc received from ball 2 on display 110.

[0056] Fig. 10 is a flowchart showing an example of the flow of the classification process performed by the processor 202 executing the analysis program Pga2 of Fig. 8. The flowchart shown in Fig. 10 is a flowchart in which S101, S108, and S109 in Fig. 7 are replaced with S121, S128, and S129, respectively.

[0057] 10, after acquiring pitch data from the measurement sensor in S121, processor 202 performs S102 to S107 in the same manner as in embodiment 1. In S128, processor 202 transmits classification result Rc including the pitcher name and pitch type of the label attached to the correct answer candidate data to information processing terminal 10B, and ends processing. In S129, processor 202 transmits classification result Rc including the fact that the pitcher corresponding to the pitch characteristic data is not registered to information processing terminal 10B, and ends processing.

[0058] According to the analysis system of embodiment 2, the data transmitted from sensor device 20B to the information processing terminal for each pitch is the classification result Rc, which has a smaller amount of data than the pitch data, which is time-series data of physical quantities, and therefore the communication load between ball 2 and the analysis device for each pitch can be reduced.

[0059] As described above, the analysis method, analysis device, analysis system, and analysis program according to the second embodiment can improve the efficiency of analysis of data acquired from a sensor fixed to a ball.

[0060] [Embodiment 3] In the first and second embodiments, a configuration has been described in which a rule-based similarity calculation is used in the classification process based on the pitch identification dataset. In the third embodiment, a configuration will be described in which the classification process uses a classification model that has been trained by machine learning using the pitch identification dataset as training data.

[0061] Fig. 11 is a block diagram showing an example of a hardware configuration of information processing terminal 10C according to embodiment 3. In the configuration of information processing terminal 10C, analysis program Pga in memory 104 in Fig. 4 is replaced with Pga3, and a machine learning program Pgs and classification model Mc are added to memory 104. Other than this, the configuration of information processing terminal 10C is similar to the configuration of information processing terminal 10, and therefore description of the similar configuration will not be repeated.

[0062] The processor 102 executing the machine learning program Pgs performs machine learning on the classification model Mc using the pitch identification dataset Sdp as learning data. In the machine learning, for example, supervised learning is performed on each of the multiple pitch identification data included in the pitch identification dataset Sdp, with the labels attached to the multiple pitch identification data being treated as correct classes. The classification model Mc includes, for example, a convolutional neural network (CNN) or a recurrent neural network (RNN). The classification model Mc may be stored in the recording medium 115. Note that the classification model Mc may be considered to have been trained in a learning device different from the information processing terminal 10C, and may be input to the information processing terminal 10C via the recording medium 115.

[0063] Fig. 12 is a flowchart showing an example of the flow of the classification process performed by the processor 202 that executes the analysis program Pga3 of Fig. 11. S101 to S103 shown in Fig. 12 correspond to S101 to S103 in Fig. 7, respectively.

[0064] 12, processor 202 executes S101 to S103 in the same manner as in embodiment 1, and then in S134 classifies the pitch characteristic data into one of a plurality of classes using classification model Mc, and proceeds to process S135. In S135, processor 202 saves pitch data Dm associated with the pitcher name and pitch type corresponding to the class to which the pitch characteristic data belongs in memory 104 as an element of pitch data set Sdm, and proceeds to process S136. In S136, processor 202 displays on display 110 the pitcher name and pitch type corresponding to the class to which the pitch characteristic data belongs, and ends the process.

[0065] According to the analysis system of the third embodiment, the classification model Mc can be adapted to the pitch identification dataset Sdp by machine learning, and therefore, by increasing the number of pitch identification data included in the pitch identification dataset Sdp, it is possible to improve the accuracy of the classification calculation while maintaining the time required for the classification calculation. Note that, as in the second embodiment, a classification calculation using the classification model Mc may be performed on the ball 2.

[0066] As described above, the analysis method, analysis device, analysis system, and analysis program according to the third embodiment can improve the efficiency of analysis of data acquired from a sensor fixed to a ball.

[0067] In the first to third embodiments, the ball 2 is a baseball ball. The ball 2 used in the configuration according to the embodiments may be a ball for any ball game, such as a ball used in softball, handball, American football, or water polo.

[0068] [Note] The above-described embodiment of the present disclosure includes the following configuration.

[0069] <Configuration 1> acquiring motion data from at least one sensor affixed to the ball, the motion data including at least one physical quantity characterizing at least one of a translational motion and a rotational motion of the ball associated with a particular motion of a subject using the ball; and determining which of the at least one label the motion data corresponds to by a classification calculation applied to at least one identification data each having at least one label attached thereto; An analysis method, wherein each of the at least one label includes information capable of identifying a person who performed the specific action corresponding to the identification data to which the label is attached.

[0070] <Configuration 2> The analysis method according to configuration 1, wherein the classification calculation identifies a correct answer candidate data among the at least one identification data that has a maximum similarity with the motion feature data extracted from the motion data.

[0071] <Configuration 3> The analysis method according to configuration 2, wherein the step of determining which of the at least one label the action data corresponds to determines that the action data corresponds to the label assigned to the correct answer candidate data when the similarity between the correct answer candidate data and the action feature data is greater than a discrimination threshold.

[0072] <Configuration 4> 4. The analysis method according to claim 2, wherein the similarity is calculated by dynamic time warping (DTW).

[0073] <Component 5> The analysis method according to configuration 1, wherein the classification calculation classifies the motion data into one of at least one class corresponding to each of the at least one label using a classification model that has been trained by machine learning using the at least one identification data as training data.

[0074] <Component 6> The analysis method according to any one of configurations 2 to 5, wherein the motion feature data is generated from the motion data from a specific time when the subject released the ball to a time that is a reference time interval prior to the specific time.

[0075] <Component 7> 7. The analysis method according to any one of configurations 1 to 6, wherein the at least one sensor includes at least one of an acceleration sensor, a gyro sensor, and a geomagnetic sensor.

[0076] <Component 8> A processing unit that acquires motion data from at least one sensor fixed to the ball, the motion data including at least one physical quantity that characterizes at least one of a translational motion and a rotational motion of the ball associated with a specific motion of a subject using the ball; A storage unit in which the operation data is stored, The processing unit determines which of the at least one label the motion data corresponds to by a classification calculation used for at least one identification data each having at least one label attached thereto; An analysis device, wherein each of the at least one label includes information capable of identifying a person who performed the specific action corresponding to the identification data to which the label is attached.

[0077] <Component 9> At least one sensor that measures motion data including at least one physical quantity that characterizes at least one of a translational motion and a rotational motion of the ball associated with a particular motion of a subject using the ball; a processing unit for determining which of the at least one label the motion data corresponds to by a classification calculation applied to at least one identification data each having at least one label attached thereto; An analysis device, wherein each of the at least one label includes information capable of identifying a person who performed the specific action corresponding to the identification data to which the label is attached.

[0078] <Component 10> At least one sensor fixed to the ball for measuring motion data including at least one physical quantity characterizing at least one of a translational motion and a rotational motion of the ball associated with a particular motion of a subject using the ball; and an analysis device for determining which of the at least one label the motion data corresponds to by a classification calculation applied to at least one identification data each having at least one label attached thereto; An analysis system, wherein each of the at least one label includes information capable of identifying a person who performed the specific action corresponding to the identification data to which the label is attached.

[0079] <Component 11> An analysis program that, when executed by at least one processor, causes the at least one processor to execute a plurality of operations, the plurality of operations including: acquiring motion data from at least one sensor affixed to the ball, the motion data including at least one physical quantity characterizing at least one of a translational motion and a rotational motion of the ball associated with a particular motion of a subject using the ball; and determining which of the at least one label the motion data corresponds to by a classification calculation applied to at least one identification data each having at least one label attached thereto; An analysis program, wherein each of the at least one label includes information capable of identifying a person who performed the specific action corresponding to the identification data to which the label is attached.

[0080] The embodiments disclosed herein are intended to be combined appropriately within a range that is not inconsistent. The embodiments disclosed herein should be considered to be illustrative and not restrictive in all respects. The scope of the present invention is defined by the claims, not the above description, and is intended to include all modifications within the scope and meaning equivalent to the claims. [Explanation of symbols]

[0081] 2 ball, 5 subject, 6 catcher, 10, 10B, 10C information processing terminal, 20, 20B sensor device, 62 capsule, 64 silicone gel, 102, 202 processor, 104, 204 memory, 106 touch panel, 108 button, 110 display, 112 wireless communication unit, 113 communication antenna, 114 memory interface, 115 recording medium, 116 speaker, 118 microphone, 120, 210 communication interface, 205 measurement sensor, 212 storage battery, 1000 analysis system, Df pitch feature data, Dm pitch data, Dp11 to Dp13, Dp21 to Dp23 pitch identification data, Lb11 to Lb13, Lb21 to Lb23 label, Mc classification model, Pga, Pga2, Pga3 Analysis program, Pgm measurement program, Pgs machine learning program, Pqc synthetic physical quantity, Pqx~Pqz physical quantities, Rc classification result, Sdm pitch dataset, Sdp pitch identification dataset, Tr trajectory.

Claims

1. acquiring motion data from at least one sensor affixed to a ball, the motion data including at least one physical quantity characterizing at least one of a translational motion and a rotational motion of the ball associated with a particular motion of a subject using the ball; and determining which of the at least one label the motion data corresponds to by a classification calculation applied to at least one identification data each having at least one label attached thereto; An analysis method, wherein each of the at least one label includes information capable of identifying a person who performed the specific action corresponding to the identification data to which the label is attached.

2. The analysis method according to claim 1 , wherein the classification calculation identifies, from among the at least one identification data, a correct answer candidate data having a maximum similarity with the motion feature data extracted from the motion data.

3. 3. The analysis method according to claim 2, wherein the step of determining which of the at least one label the action data corresponds to determines that the action data corresponds to the label assigned to the correct answer candidate data if the similarity between the correct answer candidate data and the action feature data is greater than a discrimination threshold.

4. The analysis method according to claim 2 , wherein the similarity is calculated by dynamic time warping (DTW).

5. The analysis method according to claim 1 , wherein the classification calculation classifies the motion data into one of at least one class corresponding to each of the at least one label using a classification model that has been trained by machine learning using the at least one identification data as training data.

6. The analysis method according to claim 2 , wherein the motion characteristic data is generated from the motion data from a specific time when the subject released the ball to a time preceding the specific time by a reference time interval.

7. The analysis method according to any one of claims 1 to 6, wherein the at least one sensor includes at least one of an acceleration sensor, a gyro sensor, and a geomagnetic sensor.

8. A processing unit that acquires motion data from at least one sensor fixed to the ball, the motion data including at least one physical quantity that characterizes at least one of a translational motion and a rotational motion of the ball associated with a specific motion of a subject using the ball; A storage unit in which the operation data is stored, The processing unit determines which of the at least one label the motion data corresponds to by a classification calculation used for at least one identification data each having at least one label attached thereto; An analysis device, wherein each of the at least one label includes information capable of identifying a person who performed the specific action corresponding to the identification data to which the label is attached.

9. at least one sensor that measures motion data including at least one physical quantity that characterizes at least one of a translational motion and a rotational motion of the ball associated with a particular motion of a subject using the ball; a processing unit for determining which of the at least one label the motion data corresponds to by a classification calculation applied to at least one identification data each having at least one label attached thereto; An analysis device, wherein each of the at least one label includes information capable of identifying a person who performed the specific action corresponding to the identification data to which the label is attached.

10. At least one sensor fixed to the ball for measuring motion data including at least one physical quantity characterizing at least one of a translational motion and a rotational motion of the ball associated with a particular motion of a subject using the ball; and an analysis device for determining which of the at least one label the motion data corresponds to by a classification calculation applied to at least one identification data each having at least one label attached thereto; An analysis system, wherein each of the at least one label includes information capable of identifying a person who performed the specific action corresponding to the identification data to which the label is attached.

11. An analysis program that, when executed by at least one processor, causes the at least one processor to execute a plurality of operations, the plurality of operations including: acquiring motion data from at least one sensor affixed to a ball, the motion data including at least one physical quantity characterizing at least one of a translational motion and a rotational motion of the ball associated with a particular motion of a subject using the ball; and determining which of the at least one label the motion data corresponds to by a classification calculation applied to at least one identification data each having at least one label attached thereto; An analysis program, wherein each of the at least one label includes information capable of identifying a person who performed the specific action corresponding to the identification data to which the label is attached.

Images