Classifying Myoelectric Signals using Minimum-distance Optimum Bayes Methods

A.O. Abdul Salam and Y. Al-Assaf (United Arab Emirates)


Myoelectric Signals, Classification Methods, Minimum Distance and Optimum Bayes Classifiers.


The classification of Myoelectric signal using the Minimum-Distance and the Optimum Bayes methods is considered in this paper, the ultimate goal of which is to improve the Myoelectric system control performance. Certain statistical features are to be specified to render the application of such classification methods more feasible. These features might assume different forms and the selection of which highly influences the general computational complexity and the desired classification accuracy. In this work we have adopted the statistical mean, applied to the input signal, as for the decision making computation implemented by both classifiers. A moving average window length is properly configured to provide batches of the mean value for the incoming signal. The classifiers then are to assign these sampled mean values to their relevant patterns where matching by the minimum statistical distances and the probability distributions have been achieved. It is shown that under certain conditions the performance of the minimum distance classifier would be equivalent to the optimum Bayes classifier. The analysis and performance of these classification methods are presented and verified by using simulation exercises computed on the real trained signal as obtained from the two channels of an upper arm elbow movement.

Important Links:

Go Back