Design of a Tennis Training Assistance System Under Sensory Recognition Technology

With the gradual scientific development of current tennis training methods, tennis training assistance systems play an important role in improving training effectiveness and enhancing training efficiency. The traditional tennis training assistance system still has problems such as low accuracy of motion recognition and long response time. This article aims to use somatosensory recognition technology to design a tennis training assistance system to better meet the training needs of students and trainers. The article first introduces the overall structure of the tennis training assistance system. Then, this article collected data using KinectV2 cameras. Afterwards, it utilizes multi-target tracking to obtain joint position information, and then uses the DeepSort multi-objective algorithm to collect joint position information. Finally, this article uses a human pose map and a stacked model to achieve real-time recognition of continuous actions in the system. To test the effectiveness of somatosensory recognition technology in tennis training assistance systems, this article compared it with traditional systems. The research results show that for dataset 10, the action recognition accuracy of the paper's system is 95.4%, and the response time is only 0.013 ms. The accuracy of action recognition in traditional systems is only 80.8%, and the response time reaches 0.052 ms. The results indicate that the tennis training assistance system under the proposed method has the best accuracy in motion recognition and the shortest system response time. This study highlights the important impact of somatosensory recognition technology on the accuracy and response time of action recognition in tennis training assistance systems, providing more basis for promoting the widespread application of somatosensory recognition technology in sports training assistance systems.