Deep Learning-Based Tennis Shot Type Prediction Using Gait Analysis

This paper proposes a real-time tennis shot prediction system based on computer vision and gait analysis. This method uses the Position Pipeline to extract the tennis player's position markers from the webcam sequence. These positional signals are then used to calculate the player's gait characteristics, such as step length and cadence. A machine learning model based on gait measurements is then used to predict the type of shot the player will make. The expected recording style is displayed in real time on the webcam video. This technique was tested using a dataset of tennis players hitting different shots. The algorithm accurately predicts shot patterns 95% of the time. Tennis players who follow the recommended method can improve both their play and technique. It can also be used to develop new training programs and evaluate the effectiveness of different training philosophies.