Fusion classification of stroke patients' biosignals by weighted cross-validation-based feature selection (W-CVFS) method

A multi-source information fusion-based disease class classification of stroke patients was implemented to address the low accuracy pure input motion and electromyographic signals. sEMG sensor MYO arm ring wearable wireless Shimmer were used as data acquisition devices. The Butterworth high-pass filter filtering envelope thresholding method detected activity segment. Detection FIR using window function remove interference from signal. weighted cross-validation-based feature selection (W-CVFS) is proposed for fusion selection. top 10 features selected by W-CVFS all 18 are deep neural network training testing, result 79.17%, which better than existing mRMR (66.67%) ILFS (62.50%). classificatip9uon 95.385%, higher that a single signal or sEMG. experiments showed can retain have more influence on results improve rehabilitation model patients.