A Study on Multi-motion Pattern Recognition of EMG Based on Genetic Algorithm

The traditional neural network has the uncertain shortcomings on structures and the genetic algorithm has the optimization ability. So the improved adaptive genetic algorithm is used to optimize the numbers and value of hidden nodes of neural network. And then LMS algorithm is used to optimize the weight and threshold value of the neural network. The nueral network is trainned by much groups of training samples and the ultimate neural network system is got. Afterwards, using the improved power spectrum K value method to extract the characteristic value of the collected surface emg signals;Finally, the characteristic values of six kinds of hand motions is input into RBF neural network classifier for pattern recognition. This trial gets good results and 6 kinds of action recognition rate to 75%.