EMG pattern recognition based on evidence accumulation for prosthesis control

The authors present an EMG pattern recognition method to identify motion command for the control of a prosthetic arm by evidence accumulation with multiple parameters. The adaptive cepstrum coefficients which the authors propose in this paper, integral absolute value (IAV), difference absolute mean value (DAMV), variance and autoregressive (AR) model coefficients, are extracted as parameters by probabilistic and stochastic models. Pattern recognition is carried out through an evidence accumulation procedure with the distances measured with reference parameters. A fuzzy mapping function is designed to transform the distances. A recognition comparison test is provided to show the superiority of the suggested recognition method. A separability comparison test is also provided to evaluate the feasibility of the adaptive cepstrum coefficients extracted by the proposed approach