Deep Learning Algorithms for Recognition of Badminton Strokes: A Study Using SDNN, RNN, and RNN-GRU Models with Off-Court Video Capture

In this study, three deep learning algorithms were used to classify three badminton strokes: forehand drive, forehand clear, and smash. The traditional manual methods of stroke recognition were time-consuming and error-prone, while the deep learning algorithms provided a faster and more accurate analysis of the various strokes in real time. To capture the players’ motions from off-court angles, the study recommended a video capture method. The research evaluated the performance of several deep learning models, including simple dense neural network (SDNN), recurrent neural network (RNN), and RNN with an additional gated recurrent unit (GRU) layer (RNN-GRU), using OpenCV programming and the MediaPipe keypoints library for feature extraction. The dataset was split into an 80:20 ratio for training and validation, and 300 shot videos were collected for each stroke to evaluate the accuracy and losses of each model. Coaches and players can benefit from the study's outcomes by gaining more objective insights into the game, allowing them to develop more effective strategies.