-In this paper, a discrimination system, using a neural network for electromyographic (EMG) externally controlled upper extremity prostheses is proposed. In this system, the Artificial Neural Network (ANN) is used to learn the relation between the power spectrum of EMG signal analyzed by Fast Fourier Transform (FFT) and the performance desired by handicapped people. The Neural Network can discriminate 7 performances of the EMG signals simultaneously. In order to prove the effectiveness of this system, experiments for discriminating the 7 arm performances of a healthy 23 year-old man, were carried out. For real-time operation, a Digital Signal Processor (ADSP-21061) operates over the resulting set of weights and maps the incoming signal to the stimuli control domain. Results show a highly accurate discrimination of the control signal over interference patterns. Keywords -Upper limb prosthesis, Feedforward artificial neural network, electromyography signal, correlational analysis, data engineering. I.INTRODUCTION There has always been a goal in biomedical research to design a myoelectric (EMG) controlled upper limb prosthesis, which can be used for amputated people.The application of EMG controlled prosthesis using neuromuscular stimulation on a muscle mostly depends on the successful discrimination of the myoelectrical signal by which the control over the impeded movement shall be performed.The Universidad Nacional de Colombia, has developed an upper limb prosthesis which successfully uses EMG signals from biceps and triceps to enable amputees to open and close their hands. However, the issues on palm flexion and dorsiflexion still remain unsolved.Accordingly, there has been a lot of research trying to find ways to add two motions, wrist pronation and supination, to the capabilities of the prosthesis to create an upper arm prosthesis beyond the two degrees of freedom enabled by the arm made at Universidad Nacional. Approaches towards this issue have included signal filtering, spectral analysis and pattern recognition.Most of the research studies carried out up to now have processed signals coming from the biceps and triceps. However, we presume that additional information might be extractable from the signals obtained by electrodes at special location in the upper extremity. Although these electrodes are located at places where there is relatively little EMG activity, which is an electrical manifestation of neuromuscular activities associated with a contraction muscle. It is expected that signals from these places reflect the spatial nature of the EMG signal. This approach has been in several research studies used [1,2]. II.PROCEDURE The purpose of this research is to develop a hardware implementation of a neural network. The neural network is used in this system to learn the relation between the power spectrum of EMG s ignal, analyzed by FFT, and the performance desired by handicapped people. The Neural Network can discriminate 7 performances of the EMG signals simultaneously; the myoelectrical features...

No abstract