Basketball as a high-intensity confrontational sport, athletes’ psychological quality directly affects competitive performance. In this study, we propose a multi-physiological signal fusion detection method based on support vector machine algorithm to address the difficulty of detecting anxiety in the same-court confrontation of basketball players. METHODS: Electroencephalography (EEG), electromyography (EMG) and electrodermal (EDA) signals were collected from 10 basketball players, and 1000 sets of data samples were obtained by watching different emotion-evoking videos. The Relief algorithm was utilized for feature selection to reduce the original 100- dimensional features to key features, and combined with support vector machine and least squares support vector machine for classification and recognition. RESULTS: The Relief-SVM algorithm reduced the EEG EMG fusion features from 30 to 15 dimensions, and the recognition rate of anxiety reached 83.355%, which was 9.121% and 9.357% higher than that of EEG and EMG alone, respectively. The three-signal fusion of ECG, EMG, and SCL improved the recognition accuracy from 79.58% to 92.65% after feature selection by SBFS. CONCLUSION: The multi-physiological signal fusion method effectively improves the anxiety detection accuracy, and the support vector machine algorithm performs well in processing small-sample high-dimensional data. The method can realize realtime monitoring of basketball players’ anxiety and provide technical support for scientific adjustment of training tasks.
