Surface electromyography (SEMG) has recently emerged as a nonconventional human computer interface in various fields, such as prosthetic controls, gaming and rehabilitation of upper extremities. Electrical signals generated by muscle movements are translated into commands in gaming and control applications, providing a new interface between human and external devices. In rehabilitation alone or game based rehabilitation, myoelectric signals are used for guiding the rehabilitation process or assessing the progress. Most myoelectric control applications rely on correct classification rate and recognition accuracy of movements generating the myoelectric signal set. In this study, a multi-channel SEMG system is developed to recognize seven different upper limb movements. This study also provides a comparison among three classifiers for recognition of wrist movements. The system offers 94.1% accuracy for the seven-class classification problem, 97.2% accuracy for the six-class classification problem and 98.6% accuracy for the four-class classification problem.
