Improvement of the Guilin-Hills Myoelectric Signal Classifier for Prostheses Control under the Influence of Fatigue

S. Herrmann and K. Buchenrieder (Germany)


Signal Identification, GuilinHills Classifier, Kernelbased classifier, DaviesBouldin Index, MuscleFatigue, Prosthe ses control


Control of upper limb prostheses can be realized via the detection, processing and classification of myoelectric signals. Pattern recognition techniques enable users to interact with computers by gestures or hand movements. The prime application of our work are hand prostheses con trolled by myoelectric signals. The developed techniques and the work is not limited to prostheses control and is equally applicable for the control of exosceletons, games and other computer devices. With our established classification algorithm, the Guilin-Hills Selection Method, we can distinguish up to nine hand positions. The method is based on statistical cluster analysis and is superior to the previously employed neural-net approach. Since muscle fatigue affects the accuracy of the classification algorithms, the probability density functions of time-domain features are no longer normally distributed as previously assumed. In this paper, we present an enhancement of our classification algorithm yielding better classification results under the influence of fatigue. The new method is a kernel-based approximation of the features probability density function, leading to fewer class overlap and better classification results.

Important Links:

Go Back