Advancing Tennis Analytics: A Hybrid CNN-SVM Approach for Serve Action Classification

The study seeks to determine the effect of blending the eradicate CNN-SVM model on accurately identifying tennis serve shot movements. The model utilizes a large dataset of high-definition video recordings of professional tennis serves to categorize serve motions into five distinct types: flattening, slicing, front, under the arm, and regular serves. The CNN-SVM method consists of an attention mechanism and the SVM classifier that produces high precision in identifying the frame type and the video clip. The analyses of the model indicate that the model possesses a Macro Average, recall, and F1-Score of ${91.96 \%}$ (yt.), ${92.28 \%,}$ and ${91.95 \%,}$, respectively. The Precision-Recall-Trapezoidal results indicate that the model is highly accurate, with 92.44% accuracy, 92.01% recall, and an F1-Score of 92.06%. At the same time, the Micro Average is registering a huge score of 92.01% in all three ranges in its criterion. The model shows how good it is in correctly diagnosing serve shot motions even from different situations, indicating that it can be used effectively as a crucial instructional and analytical tool in tennis. The proposed study will also involve data acquisition from non-professional players in the short term. Thus improving the model potential in a wide range of situations. Further, the research aims to be at the forefront of developing a real-time serve shot action detection method that can assist coaches in training players professionally and supporting the analysis of live matches.