Footstep Localization on Badminton Court with Vibration Signals: A Pilot Study

Abstract

Listen

Translate article

Assessment of a badminton player’s footwork is critical. However, the automated footwork assessment method is lacking. The purpose of the study is to investigate how seismographs can be used to collect vibration signals to locate the footsteps of a player on the badminton court. Four seismographs are positioned at the four corners of the badminton court to acquire the vibration signals of two players’ footsteps. After signal preprocessing, multiple features are extracted from the preprocessed vibration signals, including the maximum amplitude AMPmax, the index of the maximum amplitude INDmax, and area under the waveform of the signal AUW. The latter two features are selected to predict the localization of the footstep after correlation analysis of the features. A multilayer perceptron (MLP) and a support vector machine (SVM) are trained to combine all the features to predict the locations of the footsteps into one of the eighteen zones of the badminton court. Six-fold and leave-one-out (LOO) cross-validations are used to estimate the accuracy of the localization method. All three extracted features are correlated with the footstep location, and AMPmax and AUW are highly correlated. Both the six-fold and LOO cross-validations indicate that the overall accuracy is 98–99%, using either the MLP or the SVM. These promising results indicate that the proposed approach has a potential to trace badminton player’s footwork accurately and future studies are warranted to investigate the utilities of the vibration signals in badminton player’s footwork assessment.