Signal Processing And Recognition System Oriented To Multi-Movements Myoelectric Prosthetic Hand

There is an issue with applying artificial intelligence, in a timely manner, tohumankind, particularly to the disabled. In this paper, we propose a surfaceelectromyography signal recognition system for prosthesis application for thehand disabled, which achieves simplicity, reliability, accuracy, and shortresponse time by increasing the performance of each part of the system and thecoordination between the interconnected components.Firstly, a surface electromyography signal acquisition system was designedon the basis of the cost and processing speed.Secondly, a method called ‘extreme-value’ was carried out on the originalsignal containing five continuous movements, by separating the signal intoisolated segments representing different postures, which made the application ofthe system for daily use possible.Thirdly, on the basis of time-domain, chaos, and time-frequency domainanalysis methods, four features, namely, the average amplitude, fractal dimension,maximum Lyapunov exponent, and wavelet coefficient were extracted from fourpossible arm locations that to be determined, Further, the amplitude average fromthe extensor digitorum and the wavelet coefficient from the flexor pollicis longuswere determined as the final features after comparing the clustering effects of theextracted features.Finally, a new strategy for classifying the different postures based on a backpropagation neural network was introduced, which obtains an average systemaccuracy of five continuous movements by82.77%.