AMIRAHMAD KHANIAN - Islamic azad university mashhad branch, biomedical engineering group, mashhad, Iran.
PARVANEH TAVAKOLI
SAEED TOOSIZADEH
HAMIDREZA KOBRAVI

In this paper, a nonlinear classification method based on the neural networks is presented for detection of different finger movements. Forearm muscles SEMG signal during different the finger movements were recorded. Then, the intended movement is detected through a presented signalanalyses method. In the proposed methodology, the preprocessed SEMG signals were segmented.Then, the extracted features related to each segment were used to classify the intended movement. Theneural networks were used as the nonlinear classifiers to discriminate the 11 different movements. Theperformance of neural networks was compared with the performance of a linear classifier called SVM. According to the results, the performance of neural network in offline and online mode are 90.57% and 98.14%, respectively, and performance of SVM classifier in offline and mode is 83.58%.

electromyography, neural network, support vector machine (SVM), hand prostheses