Badminton Strokes Recognition using Inertial Sensor and Machine Learning Approach

Sport activity recognition using inertial sensor has been one of the significant researches in improving the performance analysis of sport technology studies. For example, in badminton the classification of each stroke is significant in order to produce notational analysis so that the performance of the player can be further assessed by the coaches and sport analysts. This paper aimed to recognize badminton strokes type from the player with the use of inertial sensor and machine learning techniques. Machine learning techniques which consist of pre-processing, features extraction, classification, and performance evaluation from several classifiers such as Decision Tree, k-Nearest Neighbor, and Support Vector Machine (SVM) have been applied in this study. This proposed technique can perform simplicity in building the automated model and produced fast algorithm using MATLAB. Generally, the dataset is collected from ten players wearing inertial sensors at chest and wrist while performing badminton strokes including clear, drive, drop shot, net shot, and smash activities. The result indicates that Support Vector Machine classifier with Cubic SVM produced the highest classification accuracy which is 83.4% compared to other classifiers. It is envisaged that this proposed technique can be very useful in constructing simple, fast, and accurate model for sport activity recognition studies.