An image segmentation-based sports technology action error recognition method is proposed to address the issues where traditional sports behavior error recognition algorithms cannot effectively suppress spatial background information, lack information interaction between networks, cannot model global temporal correlations, and have large recognition errors. This paper establishes a feature detection model for the table tennis serve action image using 30 image frames sampled at a particular sampling rate. In addition, the segmentation attention network is used to extract the image’s deep features, and a feature fusion mechanism is introduced to improve the information interaction between different convolutional layers. The bidirectional LSTM (Bi-LSTM) network is then fed the deep features to construct the long-term information of the table tennis serve action. The model is used to implement the error recognition of the table tennis serve video collection image. The simulation results demonstrate that the method employed for table tennis service error recognition has a high degree of accuracy, and that the performance reliability of error recognition is enhanced. In addition, this method also has a high recognition accuracy in the move and UCF101 datasets.
Guo et al. (Wed,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: