Surface myoelectric signals (MES) are used as effective input for the control of upper arm prostheses (artificial limbs). Initially, measurements using amplitudes were widely used to classify the parameters of MES, but a very simple approach used for the three-state system for each MES control site imposed a practical limit (e.g., hand open, rest and hand close). Therefore, in this paper we demonstrate that by using a very simple pattern classification system we can achieve high accuracy of classification. One can increase the classification accuracy by making little changes like changing the pattern recognition components used in the system. For example, different features, reduction of feature methods, and classifiers that yield a system which is improved.

Surface myoelectric signals, pattern classification, classification accuracy, feature reduction methods, classifiers, pattern recognition.