Method for Using Dynamic Trajectory Analysis System for Exerciser

The present disclosure claims priority to a Taiwan Patent Application No. 113121129 filed on Jun. 6, 2024, the disclosures of which are incorporated in their entirety by reference herein.

The present disclosure particularly relates to a method for using a dynamic trajectory analysis system for an exerciser, wherein the method is capable of real-time recording and analysis of dynamic trajectories.

In daily human life, whether for health, interests, or even profession, “exercise” plays a significant role. Ball sports, in particular, are highly popular and beloved by many. In most ball sports, an exerciser needs to invest physical strength, posture, and skills. For a professional athlete, optimizing these aspects typically involves assistance from a professional coach. However, for an average individual without professional information support and professional coaching, he/she can only rely on self-observation and adjustment during an exercise process to further optimize, for example, his/her posture and skills. This process is not only slow but also lacks scientific and data-driven basis, often leading to poor progress and diminished willingness to continue exercising for the individual exerciser. Moreover, improper adjustment can potentially cause physical injury to the exerciser. Accordingly, there are currently assistance tools and applications (APPs) available on the market that can assist the individual exerciser in optimizing his/her exercise process. For example, a baseball batting cage provides image shooting to record striking posture of a batter. For another example, a golf driving range offers a ball striking tracker that shoots swing posture of a golfer, and records direction and distance of a ball after it is struck. This information further serves as a reference for the golfer to adjust his/her striking posture. Following the above description, related detection application systems also exist. For example, Republic of China Patent No. TWI782649B, titled “Badminton Smash Measurement System and Method” discloses a system and method for measuring changes in direction and speed of a shuttlecock during a smash by a player playing badminton, providing a reference for the player to optimize his/her technique. For example, U.S. Pat. No. 11,638,853B2, titled “Augmented Cognition Methods And Apparatus For Contemporaneous Feedback In Psychomotor Learning” discloses a method for extracting a dynamic jointed skeleton model during an exercise process of an exerciser to create a scalable reference model for using in training. However, the various types of related application systems currently available on the market are all developed for professional athletes. These systems are costly and have limited widespread use. The key lies in wide variety of sports. Construction of various sport detection systems and devices combined in related applications are expensive. Additionally, data, results, and adjustment recommendations produced by these systems require further assistance from professional coaches, making these systems impractical for individual users. Accordingly, how to promote widespread adoption of this type of sport assistance system to further enhance human physical and mental health development is a pressing problem that needs to be solved at this stage.

In view of the aforementioned problem, a main object of the present disclosure is to provide a method for using a dynamic trajectory analysis system for an exerciser which is convenient and easy to operate, capable of recording dynamic posture during an exercise process, and able to quickly and in real-time generate an analysis result and a feedback information.

In order to achieve the aforementioned object, the method for using the dynamic trajectory analysis system for the exerciser in the present disclosure mainly uses integrated technologies such as computer vision, edge artificial intelligence (AI), machine learning, and human factors engineering, in conjunction with a smart mobile communication device to perform the following: automatically capturing and recording images of posture of the exerciser during the exercise process; and analyzing a sport event, sport behavior, and sport equipment in the images, and then obtaining a key information for providing assistance in optimizing the exercise process. The method uses edge AI models or cloud AI models to achieve real-time prediction of objective physical performance of objects such as a human body, sport equipment, and a ball in a real environment. A prediction result is presented in a data format and visualized manner, providing a user with a real-time feedback and analysis information.

Referring to, in the present disclosure, a dynamic trajectory analysis systemfor an exerciser mainly has a smart mobile communication devicewhich has a central processing unit (CPU), and an image capturing module, an image recognition module, an image analysis module, an event database, and a model databaseinformationally connected to the CPU. The smart mobile communication deviceis installed with an application (APP)that drives execution of the method described in the present disclosure. During execution, the present disclosure uses edge artificial intelligence (AI) models to achieve real-time prediction of objective physical performance of objects such as a human body, sport equipment, and a ball in a real environment. A prediction result is presented in a data format and visualized manner, providing a user with a real-time feedback and analysis information. Components are as follows:

(1) The CPUrefers to a CPU with computing capabilities. The CPUmay also be or further operate in conjunction with a graphics processing unit (GPU).

(2) The image capturing moduleis, for example, a camera lens.

(3) The image recognition modulecan be an algorithm or an integrated circuit module in which an algorithm is written. When the algorithm is executed in conjunction with a model database, comparison is performed on the dynamic image data according to at least one predetermined feature. Upon computation, the comparison causes recognition to be performed on images in the content of the dynamic image data to, for example, recognize a human, an object, and an action in the images. In the present disclosure, the image recognition modulein conjunction with one or more machine learning models stored in the model database(such as a sport type model, a sport behavior model, a ball equipment model, but not limited thereto), performs sport event recognition, sport behavior recognition, and sport equipment recognition on the dynamic image data.

(4) The image analysis modulecan be an algorithm or an integrated circuit module in which an algorithm is written. The algorithm is executed in conjunction with one or more machine learning models stored in the model database. Upon computation, the algorithm causes behavioral feature analysis to be performed on the content of the dynamic image data. For example, body rotation, swing direction, and swing angles during a swinging process of a baseball batter are analyzed.

(5) The event databasecan be a hard disk drive (HDD), a solid-state disk (SSD) or a virtual memory for storing digital data. In the present disclosure, the event databaseis mainly used to store a video file recorded during an exercise process of the exerciser, and various sport informations, parameters, analysis results, etc., obtained upon analysis.

(6) The model databaseis mainly used in conjunction with the image recognition moduleand the image analysis module. The model databasestores at least one machine learning model. In an embodiment of the present disclosure, the model databasestores at least a sport type model, a sport behavior model, and a ball equipment model. In addition, the model databasecan store various other machine learning models related to sports.

Following the above description, the smart mobile communication devicecan be, for example, a smart phone or a tablet computer that is convenient for the user to carry. The smart mobile communication devicehas at least one image capturing modulethat can capture and record images. In addition, the smart mobile communication deviceis installed with an application (APP)that drives execution of the method described in the present disclosure. Upon execution, the APPcan, for example, be used to drive the image capturing moduleto capture, shoot, and record images. In an embodiment, the smart mobile communication devicecan be connected to a remote server via information or communication to transmit the shot images to the remote server for storage or additional cloud computing. The remote server can perform recognition and analysis on the images and store the images. Furthermore, in one embodiment, the smart mobile communication devicecan also be a customized information device that is compact, portable, and equipped with computing, program execution, and communication transmission capabilities. For example, the smart mobile communication devicecan be an information device capable of executing the dynamic trajectory analysis systemfor the exerciser as described in the present disclosure.

Referring to, when a user wants to execute the dynamic trajectory analysis systemfor the exerciser in the present disclosure, he/she can first set up the smart mobile communication deviceat a specific location Prelative to an exerciser P (in this embodiment, a right-handed baseball batter). This specific location Pallows the image capturing moduleto shoot an entire action process using a focal distance and range of the image capturing module. Upon completion of setting up the smart mobile communication device, the user can then activate the smart mobile communication deviceand run APPto start executing the dynamic trajectory analysis systemfor the exerciser in the present disclosure. Additionally, the method S for using the dynamic trajectory analysis system for the exerciser is as follows:

(1) An image capture initiation operation S: The image capturing moduleis activated to perform image capture on a physical body of the exerciser P. This involves recording (i.e., image capture) a process of the exerciser P from preparation to completion of striking to obtain dynamic image data V.

(2) An image recognition operation S: Following the above description, the image recognition moduleperforms recognition on content of the dynamic image data V. In the present disclosure, the image recognition module, in conjunction with a first model M, performs “sport event recognition” Sto recognize what type of sport is depicted in the content of the dynamic image data V, and upon completion of the “sport event recognition” S, generate a “sport event type” A, such as baseball or golf. In the present disclosure, an example of “baseball batting” is used for illustration. As mentioned above, the first model Mrefers to the sport type modelin. In the present disclosure, the first model Mis a machine learning model for sport event recognition, which integrates key knowledge of specific sports and is trained with over ten thousand independent sports events. During the “sport event recognition” S, the first model Mcan automatically determine a type and sequence of a sport event. The image recognition module, in conjunction with a second model M, performs “sport behavior recognition” Sand “sport equipment recognition” Sto recognize sport behavior (e.g., batting) and sport equipment (e.g., a bat) in the content of the dynamic image data V, and upon completion of “sport behavior recognition” Sand “sport equipment recognition” S, generate “raw data” B. Moreover, the second model Mrefers to the sport behavior modelor the ball equipment modelin. The second model Memploys a machine vision algorithm. Through analysis of over ten thousand independent sports events and sports-specific knowledge, the second model Mcan perform processing tasks such as synchronous recognition of objects such as a human body, sport equipment, and a ball, noise elimination, and behavioral feature determination at the same time, and generate the raw data B. Following the above description, the image recognition modulefurther generates an “event to be analyzed” Sby combining the “sport event type” A and the “raw data” B. Furthermore, during this operation, based on the “sport event type” A, feedback such as “adjustment of shooting parameters” can be provided upon computation to achieve a better shooting effect.

(3) A behavioral feature data analysis operation S: Following the above description, the image analysis module, in conjunction with a third model M, subsequently further performs behavioral feature data analysis on the “event to be analyzed” S. The third model Mis a highly efficient machine learning model established through analysis of over ten thousand independent sports events and sports-specific knowledge, employing a machine vision algorithm. Referring also to, in this operation, the image analysis modulefurther analyzes the raw data of single-perspective images of the dynamic image data Vto obtain behavioral feature data C, which includes, for example, rotation of human body posture, action angles, and objects such as the sport equipment, and the ball.

(4) A physical performance prediction operation S: Following the above description, the image analysis module, in conjunction with the third model M, further predicts objective physical performance of objects such as a human body, sport equipment, and a ball in a real environment based on the behavioral feature data C.

(5) A “reading interface S” generation operation: Following the above description, the image analysis modulegenerates a reading interface Sthat arranges a prediction result in a data format and visualized manner and is presented on the smart mobile communication device, providing a user with a real-time feedback and analysis information. Preferably, the reading interface Scan be, for example, any one of the following or a combination thereof: a data list, a statistical chart list, or a dashboard-style list. The reading interface Scan also be presented by a computer device after being transmitted.

(6) As described above, the present disclosure mainly involves using the smart mobile communication device, set up at the appropriate location, to shoot the exercise process of the exerciser, thereby obtaining the dynamic image data V. Then, the smart mobile communication device, in conjunction with the various machine learning models (edge AI) stored in the model databaseon the device end, such as the sport type model, the sport behavior model, and the ball equipment modelin the embodiment of the present disclosure, performs edge computing to generate multiple key informations and data informations in real-time. Under a condition of using the smart mobile communication devicewith high-end hardware specifications, real-time information feedback can be obtained. Furthermore, based on this, the system can further predict the objective physical performance of the objects such as the human body, the sport equipment and the ball in the real environment, providing the exerciser with optimization references. Additionally, the smart mobile communication devicecan perform edge computing in conjunction with the built-in various machine learning models (edge AI). Alternatively, the smart mobile communication devicecan further collaborate with a remote server and transmits the dynamic image data Vto the remote server for additional cloud computing.

Referring to, the smart mobile communication devicefurther has a deep analysis moduleand a sports-specific knowledge database, both informationally connected to the CPU. The sports-specific knowledge databasestores a large amount of specific knowledge for various types of sports, such as optimal posture, ball striking posture, and applicable ball equipment specifications corresponding to specific types of sports. But the present disclosure is not limited thereto.

Following the above description, referring to, the deep analysis module, in conjunction with the sports-specific knowledge database, performs a deep analysis operation Son the dynamic image data Vin the event to be analyzed Safter the image recognition operation S. After the analysis is completed, a “key information extraction” operation Sis performed to extract at least one key information D from the event to be analyzed S. The key information D can be, for example, an action technique or sport performance. Furthermore, the deep analysis modulecan extract a corresponding key information based on each different specific sport to provide a more targeted analysis result for the exerciser. For example, if the sport event type A of the event to be analyzed Sis “baseball”, the key information D can be, for example, swing and striking posture, a swing angle, or a bat grip. Additionally, through the deep analysis operation Sand use of a machine vision algorithm, images in the dynamic image data Vcan be analyzed, and a high-quality sport segment can be automatically extracted. The segment is precisely extracted based on a key movement or moment within the sport event, preserving the most valuable images for analysis.

Referring to, the smart mobile communication devicefurther has a time sorting module, which is informationally connected to the CPU. Referring also to, the time sorting moduleperforms an automatic time sorting operation Safter the image recognition operation S. The automatic time sorting operation Smainly arranges video segments of dynamic image data Vin one or more events to be analyzed Sin chronological order to generate a complete and smooth “training process record” Swhich the exerciser can replay and review later. Referring also to, time sorting for presentation of the generated “training process record” Scan be further refined according to informations such as the sport event type, and rules, to present a training process that better meets needs of sport science.

Referring to, the smart mobile communication deviceis further installed with an augmented reality (AR) APP. Using combined AR technology, the smart mobile communication deviceintuitively presents a data information in the dynamic image data Vin an AR manner to the user. Referring also to, the aforementioned data information can be, for example, a trajectory of a postural change of the exerciser P during exercise, a trajectory of ball striking posture, or a trajectory of a moving ball. In the present embodiment, movement trajectories of human joints, performance of the sport equipment, etc., can be visualized, assisting the exerciser P in analyzing the action technique and the sport performance and key comments can be recorded in text. Referring also to, in the present disclosure, the AR APPcan be further combined with a built-in sensor of the smart mobile communication device, such as a G-sensor, an accelerometer sensor, or a gyroscope. When the user sets up the smart mobile communication device, the AR APPcombined with the sensor can provide a visual prompt through an AR image to assist the user in adjusting parameters such as an angle and a distance of the smart mobile communication deviceuntil the AR image meets requirements. This achieves a goal of excellently guiding the user to place the smart mobile communication devicein an optimal shooting position, ensuring that the shot images are clear and complete.

Referring to, the smart mobile communication devicecan be further informationally connected to one or more social platforms. Accordingly, the user can organize, and output or publicly share content such as a high-quality sport event video (the dynamic image data), the key information of the sport event, and visualized AR data (the data information), to the one or more social platforms. Further, through connection and application of the one or more social platforms, the user can interact with an online peer or coach by sharing the visualized movement trajectories of human joints, performance of the sport equipment, etc., analysis of the action technique, the sport performance, and the key comments recorded in text.

In summary, a method for using a dynamic trajectory analysis system for an exerciser in the present disclosure mainly uses technologies such as computer vision, edge AI, machine learning, and human factors engineering, in conjunction with a smart mobile communication device to perform the following: automatically capturing and recording images of posture of the exerciser during an exercise process; and further analyzing a sport event, sport behavior, and sport equipment in the images, and then obtaining a key information for providing assistance in optimizing the exercise process. The method uses edge AI models to achieve real-time prediction of objective physical performance of objects such as a human body, sport equipment, and a ball in a real environment. A prediction result is presented in a data format and visualized manner, providing a user with a real-time feedback and analysis information. Accordingly, the present disclosure, upon implementation, can indeed achieve an object of providing the method for using the dynamic trajectory analysis system for the exerciser which is convenient and easy to operate, capable of recording dynamic posture during the exercise process, and able to quickly and in real-time generate an analysis result and a feedback information.

The above is only the preferred embodiments of the present disclosure, and is not intended to limit the present disclosure to the forms disclosed. Any modifications, equivalent alternatives, and improvements made within the spirit and the scope of present disclosure by persons skilled in the art should be included in the scope of claims of the present disclosure.