Abstract: This study purposed and evaluates a method based on Support Vector Machine (SVM) classification of surface Electromyogram (sEMG) signals. The result shows that the best selection of time domain features gives better performance with quadratic SVM classifier. The sEMG signals are acquired from 15 healthy volunteers by placing the electrode on biceps and triceps muscles on the right arm. After the signal acquisition, pre-processing (denoising, rectification, filtering and amplitude normalization) is performed and suitable features were extracted for the classification purpose. In this research, sEMG data is acquired by two channels. PCA is used for dimension reduction of the feature vector. Higher classification accuracy is achieved by using the quadratic SVM classifier. The overall mean classification accuracy with selected time domain feature, i.e, Kurtosis, Skewness, Slop Sign Change (SSC), Mean Absolute Value (MAV), Autoregressive coefficient of the first order (AR1) is 99.04% with quadratic SVM classifier. This much accuracy can be used for designing assistive robotic device.
Vivek Ahlawat, Ritula Thakur and Yogendra Narayan, 2018. Support Vector Machine based Classification Improvement for EMG Signals using Principal Component Analysis. Journal of Engineering and Applied Sciences, 13: 6341-6345.