Advancing Table Tennis Analytics: A CNN-LSTM Framework for Precision Serve Shot Analysis

As for the current work, this research proposes a CNN-LSTM model systematized for evaluating and predicting the serve strokes in real-time table tennis. It is supposed to address the deficiency of extensive and timely sports analysis by combining the spatial FE capability of Convolutional Neural Networks (CNN) with the temporal dynamics learning function of Long Short-Term Memory (LSTM) networks. In the case of our signal, high-frame-rate video data is used to study and identify all manner of serve shots with a reasonable degree of precision. It also makes predictions that give essential indications of the strategic games. It is recommended that the suggested model be used to evaluate its performance, which was conducted with the help of the dataset containing the video recordings of table tennis serves. These recordings were captured at 120 frames per second, and each video was of the dimension $1920 \times 1080$ pixels. The presented model showcased the top performance of $\mathbf{9 2} \%$ with the precision and recall of 90 $\%$ and $\mathbf{8 5 \%}$ on the topspin serves. It proved to be $\mathbf{2 0 \%}$ better than the standard analytical methods and other independent schemes, which include CNNs and SVMs. Our study underscores the potential of the CNN-LSTM model to revolutionize sports analysis. By integrating this model, coaches, and athletes can significantly enhance their strategies and performance, thanks to the comprehensive vital information it provides. Moreover, this model improves our understanding of serve mechanics and paves the way for its application to other complex motions in sports.