Badminton Shot Recognition with LSTM Network

Deep learning has gained widespread application in action recognition and classification in sports. This paper focuses on the classification of three fundamental shots—Clear, Serve, and Smash—played in the game of badminton. For our experiments, we curated a dedicated dataset. Our approach involves the implementation of a Long Short-Term Memory (LSTM) deep learning model, trained to discern the movements of key points on a player’s body during the execution of each specific shot. The outcome of our experiments is noteworthy, demonstrating an impressive 89.49% accuracy in the classification of badminton shots.