Estimation of Cross Correlation in Classifying Neuromuscular Activities

Aydin S.

National Conference on Electrical, Electronics and Biomedical Engineering (ELECO), Bursa, Turkey, 1 - 03 December 2016, pp.474-476 identifier

  • Publication Type: Conference Paper / Full Text
  • City: Bursa
  • Country: Turkey
  • Page Numbers: pp.474-476


In the present study, linear, non-linear and statistical approaches so named Fourier Correlation, Wavelet Correlation (WC) and Pearson Correlation, respectively have been used to estimate cross-correlations between electrical muscle activities collected from two symmetric muscles and the these methods have been compared to each other with respect to classification performance. Experimental data, provided by UCI (Unv. of California Irvine), including agressive and normal measurements collected from 4 volunteers through eight surface electrodes during 10 different physical activities. The features, which are obtained by using those 3 methods for each electrode pair, are classified by using Support Vector Machines. Since, EMG series are nonstationary and the same muscle groups on right and left limbs (arm, leg) produce almost the same electrical activities, WC is found to be the best method in classifying physical muscle actions. The cross-correlation between EMG series from normal actions is much higher than that of agressive muscle actions.