JP6677320B2 - Sports motion analysis support system, method and program - Google Patents

Description

  The present invention relates to a sports motion analysis support system, a sports motion analysis support method, and a sports motion analysis support program that support motion analysis in sports.

  Motion capture is known as a technique that can be used for analyzing motion in sports.

  In addition, there is application software that displays a trajectory of a specific part of a body (for example, an ankle) based on a moving image obtained by imaging a person playing sports. In addition, there is application software that displays a difference between a model form and a sports form.

  Patent Document 1 discloses a system that superimposes and displays a skeleton image serving as a model, which is a model skeleton image, on a moving image of a learner's golf swing or the like taken by a video camera. Have been described.

JP 2005-237494 A

  The form when performing sports and the result of the operation including the form do not always correspond completely. In addition, as an example of the result of the action in the sport, a numerical value indicating the result of the sport is given. An example of a result other than the numerical value indicating the performance is an event. Specific examples of the event include, for example, items relating to the movement of equipment used in sports, such as items in which direction the ball flew after penalty kick (PK) in soccer.

  The following shows an example in which a form and a result of an operation including the form do not always completely correspond to each other. For example, even if a person's form is good, the condition may not be good and the person may not have performed well. Also, for example, even if a person's form is good, it may happen that good results are not obtained due to external factors such as rain and wind. Similarly, even if a person operates in a form in which the ball tends to fly rightward, the ball may fly leftward due to conditions, external factors, and the like.

  However, statistically, there is a tendency that a good result is easily obtained with such a form, and a tendency that a ball is likely to fly in a specific direction with such a form.

  If a user can be presented with a video that shows a behavior that includes a form that is statistically likely to produce a predetermined result from among many videos, the user can improve his or her form and the form of a competitor's player The video can be used for finding a habit or the like.

  However, in the case where image data of a moving image representing an operation including a form in which a predetermined result is likely to occur is specified from among image data of many moving images, it is assumed that individual moving images are to be confirmed one by one. Work load is very large.

  The system described in Patent Document 1 displays a skeleton image superimposed on one moving image. Therefore, when there are a large number of moving images, the user must confirm each moving image on which the skeleton images are superimposed.

  Therefore, the present invention provides a sports motion analysis support system and a sports motion analysis support system capable of easily specifying, from among many video data of a moving image, image data of a moving image representing a motion including a form in which a predetermined result is likely to occur. It is an object to provide a method and a sports motion analysis support program.

  A sports motion analysis support system according to the present invention includes a data storage unit that stores a plurality of data in which moving image data representing a series of motions in a sport and the results of the motions are associated with each other, a motion and a result corresponding to the motion. A learning unit that learns a model representing the relationship between the image data, a prediction unit that calculates a predicted value of the result based on the model for each image data, and a data storage unit based on the predicted value calculated for each image data And a selection unit for selecting a predetermined number of image data from the plurality of image data stored in the storage unit.

  In addition, the sports motion analysis support method according to the present invention includes a computer including a data storage unit that stores a plurality of pieces of data in which moving image data representing a series of motions in a sport and results of the motions are associated with each other. Learning a model representing a relationship with a result corresponding to an operation, calculating a predicted value of the result based on the model for each image data, and storing the predicted value of the result in the data storage unit based on the predicted value calculated for each image data. A predetermined number of image data is selected from a plurality of stored image data.

  Further, the sports motion analysis support program according to the present invention is a sports motion analysis support program which stores a plurality of data in which a plurality of data in which moving image data representing a series of motions in a sport are associated with the results of the motions are stored. A motion analysis support program, the computer comprising: a learning process for learning a model representing a relationship between an operation and a result corresponding to the operation; a prediction process for calculating a predicted value of the result based on the model for each image data And performing a selection process of selecting a predetermined number of image data from a plurality of image data stored in the data storage unit based on a predicted value calculated for each image data.

  ADVANTAGE OF THE INVENTION According to this invention, the image data of the moving image showing operation | movement containing the form which a predetermined result tends to produce can be easily specified from the image data of many moving images.

1 is a block diagram illustrating an example of a sports motion analysis support system according to a first embodiment of the present invention. It is a schematic diagram which shows a series of operation | movement of the person who performs long jump. It is a schematic diagram which shows the example of several data which matched image data and the result. 6 is a flowchart illustrating an example of processing progress according to the first exemplary embodiment of the present invention. FIG. 9 is an explanatory diagram illustrating an example of a predicted value obtained in step S2. It is a schematic diagram which shows the example of several data which matched image data, a grade, an external factor, and a condition. It is a schematic diagram which shows a series of operation | movement of the player who performs PK. FIG. 9 is a schematic diagram illustrating an example of a plurality of data in which image data and events are associated with each other. It is a flowchart which shows the example of a process progress of 2nd Embodiment of this invention. FIG. 9 is an explanatory diagram illustrating an example of probabilities obtained in step S12. FIG. 4 is a block diagram illustrating a configuration example when a data acquisition unit is provided. FIG. 3 is a schematic block diagram illustrating a configuration example of a computer according to each embodiment of the present invention. It is a block diagram showing the outline of the present invention.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

Embodiment 1 FIG.
In the first embodiment of the present invention, a case will be described in which a result of a sport operation is represented by a numerical value indicating a result. In the first embodiment, a long jump is described as an example of sports. Therefore, in the present embodiment, an example will be described in which the result of the long jump operation is a numerical value indicating the performance (in this example, the jump distance). However, the first embodiment is also applicable to other sports in which the result of the operation is represented by a numerical value indicating a score.

  FIG. 1 is a block diagram showing an example of a sports motion analysis support system according to the first embodiment of the present invention. The sports motion analysis support system 1 according to the first embodiment of the present invention includes a data storage unit 2, a learning unit 4, a prediction unit 5, an operation unit 8, a selection unit 6, and a display unit 7.

  The data storage unit 2 is a storage device that stores a plurality of data in which image data of a moving image representing a series of operations in a sport and results of the operations are associated with each other. In this example, in each data, moving image data representing a series of motions of a person performing a long jump and a result (jump distance) obtained as a result of the motion are associated with each other. The data storage unit 2 stores a plurality of such data.

  The manner of associating the image data with the results is not particularly limited. For example, the result may exist as data different from the image data, and the image data and the result may be associated with each other. Further, for example, the image data and the result may be associated with each other in such a manner that the result is included in the moving image of the image data (in other words, the result is displayed on the moving image). This is the same in the embodiment described later. Hereinafter, the former will be described as an example.

  FIG. 2 is a schematic diagram showing a series of actions of a person performing a long jump (hereinafter, referred to as a player). In FIG. 2, for simplicity of the drawing, for example, the same posture is shown as the posture of the athlete during the run, but the actual athlete performs a series of operations while moving his limbs and the like. Perform the operation. The athlete makes a run, jumps at the crossing board 11, and then lands. By capturing the series of operations with a video camera, moving image data representing the series of operations can be obtained. Then, data associating the image data with the result at that time becomes one data. The data storage unit 2 stores a plurality of such data in advance. The results associated with the image data are actually measured values.

  FIG. 3 is a schematic diagram illustrating an example of a plurality of data in which image data and performance (actually measured values) are associated with each other. In FIG. 3, each image data is indicated by a symbol having a serial number “#” and a number for convenience. The data storage unit 2 stores, for example, data illustrated in FIG. 3 in advance. Although FIG. 3 shows 20 data, the number of data stored in the data storage unit 2 is not limited to 20, and more data may be stored.

  When performing analysis for a specific player, data that associates the image data of a moving image obtained by imaging during the practice or competition of the specific player with the performance at that time is used as the data. What is necessary is just to store a plurality of them in the storage unit 2. Hereinafter, a case where the data storage unit 2 stores a plurality of pieces of data relating to a specific player will be described as an example. The specific player may be a player who uses the sports motion analysis support system 1 of the present invention, or a player who is instructed by a coach or the like who uses the sports motion analysis support system 1. Alternatively, the above-mentioned specific player may be a competitor for a user using the sports motion analysis support system 1 of the present invention. This is the same in other embodiments described later.

  Since the animation represents a series of actions of the player, the animation also represents the player's form.

  The learning unit 4 uses a plurality of data (for example, see FIG. 3) stored in the data storage unit 2 as teacher data, and performs a motion of a player represented by a moving image and a result (jump distance) corresponding to the motion. Is learned by machine learning (in other words, it is generated).

  When learning a model, the learning unit 4 extracts a still image from each image data of a moving image at predetermined time intervals (for example, at 0.5 second intervals). That is, the learning unit 4 extracts a set of still images from the moving image data. This set of still images represents the action and form of the player. Here, 0.5 seconds has been exemplified as an example of the above-mentioned predetermined time, but the above-mentioned predetermined time is not limited to 0.5 second.

  The learning unit 4 learns, as a model, a model in which a set of still images is used as an explanatory variable and performance is used as an objective variable. Therefore, it is possible to calculate the predicted value of the performance using the model obtained by learning and the image data of the moving image (more specifically, a set of still images extracted from the image data of the moving image). .

  Further, in the present embodiment, the model learned by the learning unit 4 can be said to be a model representing the relationship between the player's form and the performance.

  Note that the machine learning algorithm may be any algorithm that can learn a model for calculating a predicted value of performance using image data of a moving image.

  The prediction unit 5 calculates a predicted value of the performance for each image data stored in the data storage unit 2 based on the model learned by the learning unit 4.

  The prediction unit 5 may extract a set of still images from the image data when calculating the predicted value of the performance for one image data. Then, the prediction unit 5 may calculate the predicted value of the performance by applying the set of still images to the model as an explanatory variable.

  The selecting unit 6 determines whether each of the image data whose predicted values of the results calculated by the predicting unit 5 correspond to the first to upper predetermined positions and the predicted values of the results corresponding to the first to lower predetermined units. Selected image data. The values representing the upper predetermined number and the lower predetermined number are specified by the user of the sports motion analysis support system 1. It can be said that the values representing the upper predetermined number and the lower predetermined number are selection criteria when the selecting unit 6 selects the image data.

  The operation unit 8 is a user interface for the user to input values representing the upper predetermined order and the lower predetermined order to the sports motion analysis support system 1 (more specifically, the selection unit 6). In other words, the operation unit 8 is a user interface for the user to input selection criteria. The operation unit 8 is realized by, for example, an input device such as a keyboard. However, the mode of the operation unit 8 is not limited to such an input device.

  In the above description, the selecting unit 6 determines that each of the image data whose predicted value of the performance corresponds to the first to the upper predetermined order and the predicted value of the performance corresponding to the first to the lower predetermined order. The case in which the selected image data is selected has been described. The selection unit 6 selects each image data whose predicted value of the grade corresponds to the first to the upper predetermined order, and selects each image data whose predicted value corresponds to the first to the lower predetermined order. Need not be selected. Alternatively, the selection unit 6 selects each image data whose predicted value of the performance corresponds to the first to the lower predetermined order, and selects each image data whose predicted value corresponds to the first to the upper predetermined order. The data need not be selected.

  Hereinafter, the selection unit 6 determines whether each of the image data whose predicted value of the performance corresponds to the first to the upper predetermined order and the image data whose predicted value of the performance corresponds to the first to the lower predetermined order. The case where data is selected will be described as an example.

  In the case of the long jump, the higher the value of the jump distance that is the result, the better the result, and the smaller the value of the jump distance, the worse the result. Therefore, each piece of image data whose predicted value corresponds to the first highest rank to the predetermined highest rank is image data of a moving image whose performance is predicted to be good. Further, each image data whose predicted value corresponds to the first to the lower predetermined order is image data of a moving image predicted to have poor performance.

  Depending on the type of sport, the higher the score, the worse the score, and the lower the score, the better the score. In such a case, each piece of image data whose predicted value corresponds to the first to upper predetermined numbers is image data of a moving image predicted to have poor performance. In addition, each image data whose predicted value corresponds to the lower first to the lower predetermined is the image data of a moving image whose performance is predicted to be good.

  The selecting unit 6 selects the image data corresponding to the first to the highest ranks of the predicted value of the score, and the selection of the image data corresponding to the first to the lowest ranks of the predicted score. Whether the selected image data is selected may be determined in advance. That is, the selection criterion may be determined in advance. In that case, the operation unit 8 may not be provided. In addition, the selection unit 6 may be determined in advance so as to select the image data having the highest predicted value of the performance and the image data having the lowest predicted value of the performance.

  The display unit 7 displays a moving image of each image data selected by the selection unit 6. The display unit 7 displays each image data whose predicted value of the performance corresponds to the first to the upper predetermined order and each image data whose predicted value of the performance corresponds to the first to the lower predetermined order. Are displayed separately. For example, the display unit 7 displays, on the right side of the screen, a moving image of each image data in which the predicted value of the result corresponds to the first to upper predetermined positions, and the predicted value of the result is the lower to first predetermined value. May be displayed on the left side of the screen. The manner in which the two are displayed separately is not limited to the above example.

  The learning unit 4, the prediction unit 5, the selection unit 6, and the display unit 7 are realized by, for example, a CPU (Central Processing Unit) of a computer having a display device (not shown in FIG. 1). In this case, the CPU reads the sports motion analysis support program from a program recording medium such as a computer program storage device (not shown in FIG. 1), and according to the sports motion analysis support program, the learning unit 4, the prediction unit 5, and the selection unit. 6 and the display unit 7. In the display unit 7, a part that actually displays a moving image is realized by a display device, and a part that performs display control of a moving image based on image data is realized by a CPU. Further, the above computer may be a personal computer or a portable computer such as a smartphone. These points are the same in other embodiments described later.

  The sports motion analysis support system 1 may have a configuration in which two or more physically separated devices are connected by wire or wirelessly. For example, the sports motion analysis support system 1 may be realized as a system in which a portable computer such as a smartphone and a server cooperate. This is the same in other embodiments described later.

  Next, the processing progress will be described. FIG. 4 is a flowchart illustrating an example of the progress of processing according to the first embodiment of this invention. The description of the operation already described may be omitted as appropriate.

  It is assumed that the data storage unit 2 stores a plurality of data in advance. In this example, for convenience, a description will be given assuming that a plurality of data shown in FIG.

  The learning unit 4 uses the data stored in the data storage unit 2 (see FIG. 3) as teacher data and learns a model representing the relationship between the movement of the player represented by the moving image and the performance corresponding to the movement. (Step S1). In step S1, the learning unit 4 extracts a still image from each image data of a moving image at predetermined time intervals (for example, every 0.5 seconds). As a result, the set of still images is associated with the performance. Then, the learning unit 4 only needs to learn a model in which a set of still images is used as an explanatory variable and the score is used as an objective variable. As described above, this model can be said to be a model representing the relationship between a player's form and performance.

  The predicting unit 5 calculates a predicted value of the result for each image data included in the teacher data (see FIG. 3) based on the model obtained in step S1 (step S2). As described above, the prediction unit 5 extracts a set of still images from the image data and calculates the set of still images from the image data as an explanatory variable when calculating the predicted value of the performance for one image data. By doing so, the predicted value of the performance may be calculated. The prediction unit 5 may perform this processing for each image data.

  FIG. 5 is an explanatory diagram showing an example of the predicted value obtained in step S2. In this example, the description will be made on the assumption that the predicted value of the performance shown in FIG. 5 is calculated for each image data in step S2.

  Subsequently, the selection unit 6 receives an input of a selection criterion from the user via the operation unit 8 (Step S3). In the present example, the selection unit 6 receives, via the operation unit 8, an input of a value representing the upper predetermined number and the lower predetermined number. Further, in this example, a case where “2” is input as a value representing the upper predetermined number and the lower predetermined number will be described as an example.

  Subsequently, the selection unit 6 selects image data based on the predicted value of the result calculated in step S2 and the selection criterion input in step S3 (step S4). In this example, the selecting unit 6 determines that each of the image data whose predicted values of the results correspond to the first to the second highest and the predicted values of the results correspond to the first to the second lowest. Select each image data. In the example shown in FIG. 5, the top one of the predicted values is “8m95”, and this predicted value corresponds to image data # 5. The second highest order of the predicted value is “8m93”, and this predicted value corresponds to image data # 4. Further, the first lower order of the predicted value is “7m00”, and this predicted value corresponds to image data # 17. The second lowest order of the predicted value is “7m50”, and this predicted value corresponds to image data # 14.

  Therefore, in this example, the selecting unit 6 selects # 5 and # 4 as the image data whose predicted values correspond to the first to the second, and the predicted values are the first to the second two. # 17 and # 14 are selected as the image data corresponding to the first.

  Subsequently, the display unit 7 displays the moving image of the image data selected in step S4 (step S5). In this example, the display unit 7 displays the moving images # 5 and # 4 and the moving images # 17 and # 14, respectively.

  In the present embodiment, the model is generated using a plurality of combinations of the moving image and the performance. Therefore, when the predicted value of the performance is calculated by this model, it can be said that the predicted value statistically expresses the tendency of the performance according to the form. In the present embodiment, the selection unit 6 determines whether each of the image data whose predicted value corresponds to the first to the upper predetermined order and the image data whose predicted value corresponds to the first to the upper predetermined order. And the display unit 7 displays the moving image of each image data. In this example, the moving image of each image data whose predicted value corresponds to the first to the upper predetermined (the nth) is the moving image whose performance is statistically predicted to be the best. Each video up to the video that is statistically predicted to be the best. Also, the moving image of each image data corresponding to the predicted value from the lower first to the predetermined lower (n-th) image is the n-th moving image from the moving image statistically predicted to have the worst performance. Each movie up to the movie statistically predicted to be bad. Therefore, in the present embodiment, even if the number of image data is large, it is easy to identify the image data of a moving image that is statistically predicted to be good or the image data of a moving image that is statistically predicted to be bad. can do. In this embodiment, factors that do not appear in the form (external factors such as player conditions, wind, rain, etc.) are excluded, and a moving image when the performance is predicted to be statistically good, It is possible to display a moving image that is predicted to be bad. By checking these videos, the user can check the form when the performance is statistically predicted to be good or when the performance is predicted to be statistically poor without considering the conditions and external factors. You can check the form and analyze the form.

  For example, referring to FIG. 5, the image data when the actual performance is the best is # 4, but the image data indicating the form in which the statistically best performance is likely to appear is # 5. In the above example, the moving image of the image data # 4 is also displayed. However, the user particularly analyzes the image of the image data # 5 more intensively, for example, to perform the analysis for improving the form. Can be.

  In addition, # 17 shown in FIG. 5 is a moving image when the actual performance is the third worst. However, the predicted value of the result of # 17 is the first lowest. Therefore, it can be said that the moving image of # 17 represents a form in which the worst results are statistically likely to occur. The user can perform, for example, an analysis for improving the form by confirming the moving image of # 17 selected by the selecting unit 6.

  As described above, in the present embodiment, from the image data of many moving images, the image data of the moving image that indicates the operation including the form that easily gives a good result or the image of the moving image that includes the form that easily indicates the bad result is displayed. Data can be easily specified.

  As described above, the selection criterion may be predetermined. In addition, a selection criterion that defines selection of the first image data having the highest predicted value and the first image data having the lowest predicted value of the performance may be predetermined.

  When learning the model, the learning unit 4 uses a model that uses not only a moving image (more specifically, a set of still images extracted from the moving image) but also external factors and the condition of a player as explanatory variables. You may learn. In this case, when calculating the predicted value of the performance, the prediction unit 5 may substitute 0 for the explanatory variable of the external factor in the model or substitute a value indicating “good” for the explanatory variable of the condition. . In this case as well, in this case as well, a predicted value of the performance excluding external factors and a change in condition is obtained, and as a result, the same effect as described above is obtained. In this case, at the time of model learning, external factor and condition data are required. Therefore, data in which the image data of the moving image, the result, the external factor, and the condition are associated with each other may be stored in the multiple data storage unit 2. FIG. 6 shows an example of such data. The learning unit 4 can learn a model using not only a moving image but also a wind speed and a condition of a tailwind as explanatory variables using a plurality of data illustrated in FIG. 6 as teacher data. In the condition column shown in FIG. 6, “0” means “good” and “1” means “bad”.

Embodiment 2. FIG.
In the second embodiment of the present invention, a case will be described in which the result of a sports operation is represented by an event. In the second embodiment, a PK (penalty kick) in soccer will be described as an example. The description will be made assuming that the result of the PK operation is one of two types of events: "the ball has flew to the right" and "the ball has flew to the left." However, the second embodiment is applicable to other sports in which the result of the operation is represented as an event.

  The sports motion analysis support system according to the second embodiment of the present invention can be represented by the block diagram shown in FIG. 1 similarly to the sports motion analysis support system 1 according to the first embodiment. A second embodiment will be described. The description of the same items as in the first embodiment will be omitted as appropriate. The sports motion analysis support system 1 of the second embodiment includes a data storage unit 2, a learning unit 4, a prediction unit 5, an operation unit 8, a selection unit 6, and a display unit 7 (see FIG. 1). ).

  The data storage unit 2 is a storage device that stores a plurality of data in which image data of a moving image representing a series of operations in a sport and results of the operations are associated with each other. In this example, in each data, image data of a moving image representing a series of motions of a person who performs PK (hereinafter, referred to as a player) and an event obtained as a result of the motion (“The ball flew to the right. Or an event of “the ball has flew to the left”). The data storage unit 2 stores a plurality of such data. In this example, “right” is “right” as viewed from the player who performed the PK. The same applies to “left”.

  The event “the ball flew to the right” is represented by “1”, and the event “the ball flew to the left” is represented by “0”. That is, the result (event) of the PK operation is represented by a binary value.

  FIG. 7 is a schematic diagram showing a series of actions of a player performing PK. In FIG. 7, for simplicity of the drawing, the same posture is shown as the posture of the player performing the PK, but the actual player performs a series of operations while moving his limbs and the like. The player who performs the PK runs toward the ball 21 placed in front of the goalkeeper 22 and kicks the ball 21. Then, the ball 21 flies right or left. That is, the event “1” or the event “0” occurs. By capturing a series of actions of a player performing PK with a video camera, moving image data representing the series of actions can be obtained. Then, data in which the image data is associated with the event (“1” or “0”) at that time becomes one data. The data storage unit 2 stores a plurality of such data in advance.

  FIG. 8 is a schematic diagram illustrating an example of a plurality of data in which image data and events are associated with each other. As in FIG. 3, each image data is indicated by a symbol having a serial number “#” for convenience. Although FIG. 8 shows 20 data, the number of data stored in the data storage unit 2 is not limited to 20, and more data may be stored.

  The learning unit 4 uses a plurality of data (for example, see FIG. 8) stored in the data storage unit 2 as teacher data to represent a relationship between a motion of a player represented by a moving image and a result corresponding to the motion. Train the model by machine learning. More specifically, the learning unit 4 learns a model representing a relationship between a player's action and a probability that an event “1” occurs or a probability that an event “0” occurs due to the action.

  When learning the model, the learning unit 4 extracts a still image from each image data of the moving image at predetermined time intervals (for example, every 0.2 seconds). That is, the learning unit 4 extracts a set of still images from the moving image data. That is, the learning unit 4 extracts a set of still images from the moving image data. This operation is the same as the operation of the learning unit 4 in the first embodiment. As described in the first embodiment, this set of still images represents the action and form of the player. Here, 0.2 seconds is illustrated as an example of the above-mentioned predetermined time, but the above-mentioned predetermined time is not limited to 0.2 seconds.

  The learning unit 4 learns a model in which a set of still images is used as an explanatory variable and a probability that an event “1” occurs or a probability that an event “0” occurs is a target variable. Therefore, using the model obtained by learning and the moving image data (more specifically, a set of still images extracted from the moving image data), the probability that the event “1” occurs or the event “0” Can be calculated. Here, a case will be described as an example where the probability that the event “1” occurs or the probability that the event “0” occurs is represented by one objective variable that can take a value in the range of 0 to 1. In this case, if the value of the objective variable is greater than 0.5, the value represents the probability that an event “1” (that is, the event “the ball flies right”) will occur, and the value will be 1 The closer the value is, the higher the probability of occurrence of the event “1”, and the closer the value is to 0.5, the lower the probability of occurrence of the event “1”. If the value of the objective variable is smaller than 0.5, the value represents the probability that an event “0” (that is, the event “the ball flies to the left”) will occur, and the value becomes 0. The closer the value is, the higher the probability of occurrence of the event “0” is, and the closer the value is to 0.5, the lower the probability of occurrence of the event “0” is.

  The learning unit 4 may use, for example, logistic regression analysis as an algorithm of machine learning. For example, the learning unit 4 may learn a model that represents the relationship between the movement of the player and the probability that the event “1” occurs or the probability that the event “0” occurs due to the movement by the logistic regression analysis.

  In this example, the model learned by the learning unit 4 can be said to be a model representing the relationship between the form of the player and the probability that the event “1” occurs or the probability that the event “0” occurs.

  In the above example, if the value of one objective variable is greater than 0.5, the value represents the probability of occurrence of the event “1”, and if the value of the objective variable is less than 0.5, the value is the event “ The case where the probability of occurrence of 0 ″ is described has been described. The aspect of the model is not limited to the above example. For example, the learning unit 4 may learn a model in which the probability that an event “1” occurs and the probability that an event “0” occurs are objective variables. In this case, the probability of the occurrence of the event “1” is, for example, a value in the range of 0 to 1, and the larger the value, the higher the probability of the occurrence of the event “1”. Similarly, the probability of the occurrence of the event “0” is, for example, a value in the range of 0 to 1, and the larger the value, the higher the probability of the occurrence of the event “0”. In this case, the sum of the probability of occurrence of event “1” and the probability of occurrence of event “0” may be 1.

  Hereinafter, if the value of the objective variable is greater than 0.5, the value indicates the probability of occurrence of the event “1”, and if the value of the objective variable is less than 0.5, the probability of the occurrence of the event “0”. Is described as an example. In this case, as described above, if the value is larger than 0.5, the closer the value is to 1, the higher the probability of occurrence of the event “1”, and the closer the value is to 0.5, the more the event “1” occurs. It indicates that the probability is low. Also, if the value is smaller than 0.5, the closer the value is to 0, the higher the probability of occurrence of the event “0” is, and the closer the value is to 0.5, the lower the probability of occurrence of the event “0” is. Represents.

  The prediction unit 5 calculates a probability that an event “1” occurs or a probability that an event “0” occurs for each image data stored in the data storage unit 2 based on the model learned by the learning unit 4. The probability that the event “1” occurs or the probability that the event “0” occurs can be said to be one of the modes of the predicted value of the result corresponding to the operation.

  When calculating the probability for one image data, the prediction unit 5 may extract a set of still images from the image data. Then, the prediction unit 5 may calculate the probability (the value of the target variable) by applying the set of still images to the model as an explanatory variable.

  When image data is generated by capturing an image of a player's motion, the values (“1” and “0” in this example; see FIG. 8) representing events actually generated by the motion are, for example, nominal. It can be said that it is a measure. The value of the objective variable calculated by the prediction unit 5 based on the model represents a probability. The possible values of the objective variable are continuous values in the range of 0 to 1. Therefore, it can be said that the value of the objective variable calculated by the prediction unit 5 based on the model is, for example, an ordinal scale.

  The selection unit 6 determines whether each of the image data in which the value of the objective variable calculated by the prediction unit 5 corresponds to the first to the upper predetermined order and the value of the objective variable corresponding to the first to the lower predetermined order. Selected image data. The values representing the upper predetermined number and the lower predetermined number are specified by the user of the sports motion analysis support system 1. Further, it can be said that this value is a selection criterion when the selecting unit 6 selects the image data.

  The operation unit 8 is the same as the operation unit 8 in the first embodiment, and a description thereof will be omitted.

  In the above description, the selection unit 6 determines that each of the image data in which the value of the objective variable corresponds to the first to the upper predetermined order and the value of the objective variable corresponds to the first to the lower predetermined order. The case in which the selected image data is selected has been described. The selection unit 6 selects each image data in which the value of the target variable calculated by the prediction unit 5 corresponds to the first to upper predetermined positions, and selects the target variable from the lower first to the lower predetermined position. It is not necessary to select each image data corresponding to. Alternatively, the selecting unit 6 selects each image data in which the value of the objective variable calculated by the predicting unit 5 corresponds to the first to the lower predetermined order, and the value of the objective variable is the first to the upper predetermined It is not necessary to select the image data corresponding to the first item.

  Hereinafter, the selection unit 6 determines whether each of the image data whose objective variable values correspond to the upper first to the upper predetermined order and the image data whose objective variable values correspond to the lower first to the lower predetermined order. The case where data is selected will be described as an example.

  Note that, in this example, each piece of image data in which the value of the objective variable corresponds to the first to the top predetermined is the image data of a moving image predicted to have a high probability that the event “1” will occur. In addition, each piece of image data in which the value of the objective variable corresponds to the lower first to the lower predetermined order is moving image image data that is predicted to have a high probability that the event “0” will occur.

  Further, the selection criterion may be determined in advance. In that case, the operation unit 8 may not be provided. In addition, the selection unit 6 may be determined in advance so as to select image data in which the value of the objective variable is the first highest and image data in which the value of the objective variable is the first lowest.

  The display unit 7 displays a moving image of each image data selected by the selection unit 6. The display unit 7 displays each image data in which the value of the objective variable corresponds to the upper first to the upper predetermined number, and each image data in which the value of the objective variable corresponds to the lower first to the lower predetermined number. Display separately. For example, the display unit 7 may display a moving image of each of the former image data on the right side of the screen, and display each of the latter image data on the left side of the screen. The manner in which the two are displayed separately is not limited to the above example.

  Next, the processing progress will be described. FIG. 9 is a flowchart illustrating an example of the progress of processing according to the second embodiment of this invention. The description of the operation already described may be omitted as appropriate.

  It is assumed that the data storage unit 2 stores a plurality of data in advance. In this example, for convenience, a description will be given assuming that a plurality of data shown in FIG.

  The learning unit 4 uses the data (see FIG. 8) stored in the data storage unit 2 as teacher data, and describes the action of the player represented by the moving image and the probability of occurrence of the event “1” or the event “0” due to the action. Is learned (step S11). The learning unit 4 may learn the model by, for example, logistic regression analysis.

  In step S11, the learning unit 4 extracts a still image from each image data of a moving image at predetermined time intervals. As a result, the set of still images is associated with the actually occurring event “1” or “0”. Then, the learning unit 4 may learn a model in which a set of still images is used as an explanatory variable and a probability that an event “1” occurs or a probability that an event “0” occurs is a target variable. As described above, this model can be said to be a model representing the relationship between the form of the player and the probability that the event “1” occurs or the probability that the event “0” occurs.

  The prediction unit 5 calculates the value of the objective variable for each image data included in the teacher data (see FIG. 8) based on the model obtained in step S11 (step S12). This objective variable represents the probability that an event “1” occurs or the probability that an event “0” occurs. As described above, when calculating the probability (value of the objective variable) for one image data, the prediction unit 5 extracts a set of still images from the image data and uses the set of still images as an explanatory variable. By applying to the model, the value of the objective variable may be calculated. The prediction unit 5 may perform this processing for each image data.

  FIG. 10 is an explanatory diagram showing an example of the probabilities (values of the objective variables) obtained in step S12. Hereinafter, a description will be given assuming that the probabilities (values of the objective variables) shown in FIG. 10 are calculated for each image data in step S12.

  Subsequently, the selection unit 6 receives an input of a selection criterion from the user via the operation unit 8 (Step S13). In the present example, the selection unit 6 receives, via the operation unit 8, an input of a value representing the upper predetermined number and the lower predetermined number. Further, in this example, a case where “2” is input as a value representing the upper predetermined number and the lower predetermined number will be described as an example.

  Subsequently, the selection unit 6 selects image data based on the value of the target variable calculated in step S12 and the selection criterion input in step S13 (step S14). In the present example, the selection unit 6 determines that each of the image data in which the value of the objective variable corresponds to the first to the second highest and the value of the target variable corresponds to the first to the second lowest. Select each image data. In this example, the selecting unit 6 selects # 13 and # 1 as the respective image data in which the value of the objective variable corresponds to the first to the second highest, and the value of the objective variable starts from the first to the lowest. # 4 and # 7 are selected as the image data corresponding to the second to the lower order (see FIG. 10).

  Subsequently, the display unit 7 displays the moving image of the image data selected in step S14 (step S15). In this example, the display unit 7 displays the moving images # 13 and # 1, and the moving images # 4 and # 7, respectively.

  In the present embodiment, the model is generated using a plurality of combinations of moving images and events. Therefore, it can be said that the value calculated based on the model, which indicates the probability of occurrence of the event "1" or the probability of occurrence of the event "0", statistically expresses the tendency according to the form.

  Further, the learning unit 4 generates a model in which the probability that an event “1” occurs or the probability that an event “0” occurs is a target variable. That is, the event in the teacher data is not a continuous value, but the value of the objective variable calculated based on the model is a continuous value. Therefore, it is possible to easily rank the image data based on the value of the objective variable. Therefore, even if the number of image data is large, the image data of the moving image that is statistically predicted that the event “1” occurs (that is, the image data of the moving image that is statistically predicted that the ball will fly to the right), It is possible to easily specify image data of a moving image that is statistically predicted when the event “0” occurs (that is, image data of a moving image that is statistically predicted that a ball will fly to the left).

  By checking these videos, the user can analyze a form in which the ball is statistically predicted to fly to the right and a form in which the ball is statistically predicted to fly to the left. .

  It is also assumed that the user of the sports motion analysis support system 1 is a player who performs PK. In this case, the user stores the image data and the phenomena of his / her own moving image in the data storage unit 2 as teacher data, for example, to discover a habit of his / her own form and to eliminate the habit of the form. It can be used for improvement. Eliminating such a habit makes it difficult for the opposing team's goalkeeper to predict the direction of the ball from the form, which is advantageous for the player who performs the PK.

  Further, the user of the sports motion analysis support system 1 may be a goalkeeper. In this case, by storing the image data and the event of the moving image of the player of the opponent team in the data storage unit 2 as the teacher data, the analysis for the purpose of finding the habit of the player can be performed. If the opponent's player's habit is discovered, the goalkeeper can easily predict the direction of the ball after the PK from the opponent's player's form, which is advantageous for the goalkeeper.

  In the second embodiment, the PK of soccer is described as an example. However, data in which image data obtained by imaging a pitching operation of a baseball pitcher and events are associated with a plurality of data storage units 2 are stored. It may be stored. In this case, for example, the event may be “the ball type was straight”, “the ball type was a curve”, or “the ball type was a fork”. Then, the learning unit 4 may generate a model with the probability that the pitch is straight, the probability that the pitch is a curve, and the probability that the pitch is a fork as target variables. The selection unit 6 also includes, for example, a predetermined number of image data having a high probability that the pitch is straight, a predetermined number of image data having a high probability that the pitch is curved, and a probability that the pitch is a fork. May be selected.

  In each embodiment of the present invention, the sports motion analysis support system 1 may include a data acquisition unit that acquires data to be stored in the data storage unit 2 from outside. FIG. 11 is a block diagram illustrating a configuration example when a data acquisition unit is provided. The data storage unit 2, the learning unit 4, the prediction unit 5, the operation unit 8, the selection unit 6, and the display unit 7 are the same as those elements in the first embodiment and those elements in the second embodiment. The description is omitted.

  The data acquisition unit 9 acquires a plurality of pieces of data in which image data of a moving image representing a series of actions in sports and the results of the actions are associated with each other, and stores the data in the data storage unit 2.

  For example, it is assumed that a plurality of the above data are stored in an external device (not shown). In this case, the data acquisition unit 9 may access the device, acquire a plurality of data from the device, and store the data in the data storage unit 2. The processing after the data acquisition unit 9 stores the plurality of data in the data storage unit 2 is the same as the processing described in the first embodiment or the processing described in the second embodiment.

  The data acquisition unit 9 is realized by, for example, a CPU of a computer that operates according to a sports motion analysis support program.

  In the above description, long jump, PK in soccer, pitcher pitching in baseball, and the like have been exemplified. The sports operation to which the present invention is applied is not limited to these. For example, data in which image data of a swing operation in golf and results (flying distance) are associated with each other may be stored in the multiple data storage unit 2. Further, for example, data in which the image data of the action of the toss of the volleyball setter and the result indicating whether the ball flew to the right or left may be stored in the multiple data storage unit 2. Good. The present invention is also applicable to, for example, rugby formations and American football formations. As described above, the present invention is applicable to various sports operations.

  FIG. 12 is a schematic block diagram illustrating a configuration example of a computer according to each embodiment of the present invention. The computer 1000 includes a CPU 1001, a main storage device 1002, an auxiliary storage device 1003, an interface 1004, a display device 1005, and an input device 1006. The input device 1006 corresponds to the operation unit 8.

  The sports motion analysis support system 1 according to each embodiment of the present invention is mounted on a computer 1000. The operation of the sports motion analysis support system 1 is stored in the auxiliary storage device 1003 in the form of a program (sport motion analysis support program). The CPU 1001 reads out the program from the auxiliary storage device 1003, expands the program in the main storage device 1002, and executes the above processing according to the program.

  The auxiliary storage device 1003 is an example of a non-transitory tangible medium. Other examples of the non-transitory tangible medium include a magnetic disk, a magneto-optical disk, a CD-ROM, a DVD-ROM, and a semiconductor memory connected via the interface 1004. When the program is distributed to the computer 1000 via a communication line, the computer 1000 that has received the program may load the program into the main storage device 1002 and execute the above processing.

  Further, the program may be for realizing a part of the processing described above. Furthermore, the program may be a difference program that implements the above-described processing in combination with another program already stored in the auxiliary storage device 1003.

  A part or all of the components may be realized by a general-purpose or dedicated circuit, a processor, or a combination thereof. These may be configured by a single chip, or may be configured by a plurality of chips connected via a bus. Some or all of the components may be realized by a combination of the above-described circuit and the like and a program.

  When some or all of the components are realized by a plurality of information processing devices, circuits, and the like, the plurality of information processing devices, circuits, and the like may be centrally arranged or distributed. For example, the information processing device, the circuit, and the like may be realized as a form in which each is connected via a communication network, such as a client and server system or a cloud computing system.

  Next, the outline of the present invention will be described. FIG. 13 is a block diagram showing an outline of the present invention. The sports motion analysis support system of the present invention includes a data storage unit 2, a learning unit 4, a prediction unit 5, and a selection unit 6.

  The data storage unit 2 stores a plurality of data in which image data of a moving image representing a series of operations in a sport and results of the operations are associated with each other.

  The learning unit 4 learns a model representing a relationship between an operation and a result corresponding to the operation.

  The prediction unit 5 calculates a predicted value of the result based on the model for each image data.

  The selection unit 6 selects a predetermined number of image data from a plurality of image data stored in the data storage unit 2 based on the prediction value calculated for each image data.

  With such a configuration, it is possible to easily specify image data of a moving image representing an operation including a form in which a predetermined result is likely to occur, from image data of many moving images.

  The configuration may be such that the result of the operation associated with the image data is a numerical value indicating the performance, and the learning unit 4 learns a model representing the relationship between the operation and the numerical value indicating the performance.

  Further, the result of the operation associated with the image data may be an event, and the learning unit 4 may learn a model representing a relationship between the operation and the probability of the event occurring.

  Further, the configuration may be such that the learning unit 4 learns a model representing the relationship between the operation and the probability of occurrence of an event by logistic regression analysis.

  The present invention has been described with reference to the exemplary embodiments, but the present invention is not limited to the above exemplary embodiments. Various changes that can be understood by those skilled in the art can be made to the configuration and details of the present invention within the scope of the present invention.

Industrial applicability

  INDUSTRIAL APPLICABILITY The present invention is suitably applicable to a sports motion analysis support system that supports motion analysis in sports.

Reference Signs List 1 sports motion analysis support system 2 data storage unit 4 learning unit 5 prediction unit 6 selection unit 7 display unit 8 operation unit 9 data acquisition unit

Claims (6)

A data storage unit that stores a plurality of image data of a moving image representing a series of operations in sports, and a plurality of data in which the results of the operations are associated with each other;
A learning unit that learns a model representing a relationship between the operation and a result corresponding to the operation,
For each image data, based on the model, a prediction unit that calculates a predicted value of the result,
A selection section for selecting a predetermined number of image data from a plurality of image data stored in a data storage section based on a predicted value calculated for each of the image data. . The result of the operation associated with the image data is a numerical value indicating the grade,
The sports motion analysis support system according to claim 1, wherein the learning unit learns a model representing a relationship between the motion and a numerical value indicating a performance. The result of the operation associated with the image data is an event,
The sports motion analysis support system according to claim 1, wherein the learning unit learns a model representing a relationship between the motion and the probability of an event occurring. The sports motion analysis support system according to claim 3, wherein the learning unit learns a model representing a relationship between the motion and the probability that an event occurs by logistic regression analysis. A computer including a data storage unit that stores a plurality of pieces of data in which image data of a moving image representing a series of actions in sports and results of the actions are associated with each other.
Learning a model representing a relationship between the operation and a result corresponding to the operation,
For each of the image data, based on the model, calculate a predicted value of the result,
A sports motion analysis support method, comprising: selecting a predetermined number of image data from a plurality of image data stored in a data storage unit based on a predicted value calculated for each image data. A sports motion analysis support program mounted on a computer including a data storage unit that stores a plurality of data in which moving image data representing a series of motions in a sport and a result of the motion are associated with each other,
On the computer,
A learning process of learning a model representing a relationship between the operation and a result corresponding to the operation,
For each image data, based on the model, a prediction process of calculating a predicted value of the result, and
A sports motion analysis support program for executing a selection process of selecting a predetermined number of image data from a plurality of image data stored in a data storage unit based on a predicted value calculated for each of the image data.

Images