Recognizing Ping-Pong Motions Using Inertial Data Based on Machine Learning Classification Algorithms

With the development of Internet of Things (IoT) technology and various sensing technologies, some newer ways of perceiving people and the environment have emerged. Commercial wearable sensing devices integrate a variety of sensors that can play a significant role in motion capture and behavioral analysis. This paper proposes a solution for recognizing human motion in ping-pong using a commercial smart watch. We developed a data acquisition system based on the IoT architecture to obtain data relating to areas such as acceleration, angular velocity, and magnetic induction of the watch. Based on the features of the extracted data, experiments were performed using major machine learning classification algorithms including k-nearest neighbor, support vector machine, Naive Bayes, logistic regression, decision tree, and random forest. The results show that the random forest has the best performance, reaching a recognition rate of 97.80%. In addition, we designed a simple convolutional neural network to compare its performance in this problem. The network consists of two convolutional layers, two pooling layers, and two fully connected layers, and it uses data with no extracted features. The results show that it achieves an accuracy of 87.55%. This research can provide training assistance for amateur ping-pong players.