Electromyography signals acquisition and real-time classification

Many scholars have made research on myoelectric prosthetic hand under the new condition of more and more charity on disabled. However, the prosthetic hands on the market do not widely used for long reaction time. This paper introduces some new ideals about the real-time prosthetic. Firstly, describes a new frame of the data acquisition and processing system, analyzes the influencing factors on the state of physical fitness and the experiment environment, designs the structure of both software and hardware, and obtains high quality surface myoelectric data from this system. Secondly, designs experiments which contains continue actions and obtains the alternate data between two actions from flexor carpi radialis muscle and deep flexor muscle, the basis actions are relax, fist and stretch. Finally, extracts the general feature of the alternate data, and action patterns are classified on BP neural network, a rate of recognition for six alternate movements is 83.3% and the classification time is 2.2660s.