The study of the electromyographic signal for the control of a prosthetic hand

The design of a prosthetic device, able to reproduce the functions of a real human hand, is characterized by many challenges. The ideal prosthesis would be easy to control, comfortable to wear and aesthetically pleasing. This thesis is concerned with finding solutions to the first of the three ob- jectives mentioned above. The EMG control is the most used approach in todays prosthetic devices, because it is noninvasive compared with other methods. Its goal is to create an association between a predefined set of hand motion patterns and the corresponding EMG signals generated by the forearm muscles. In this way the classifier mounted on the prosthetic hand is able to recognize a muscle contraction and sends to the controller the command to reproduce the cor- respondent hand movement. The system designed in the current work is composed by many modules, which process the signals detected by a 3-channel EMG acquisition board. The first one is the segmentation module, which is able to understand when a muscle contraction, also referred to as signal burst, starts and ends. The second module is the features extractor, which, applied to each individual burst, extracts from it some representative parameters in the time-frequency domain, by the application of the Continuous Wavelet Transform (CWT). This generates a large matrix, which must be reduced by applying the Sin- gular Value Decomposition (SVD). The feature extractor also computes two temporal parameters which are then concatenated to the result of the SVD, to form the feature vector representing to the burst. An Artificial Neural Network is then trained to associate each feature vec- tor with the corresponding hand movement. In particular the system learns how to classify 7 different movements, with a performance of 98-100%. Eventually, the influence of some factors on the performance of the system is discussed, namely the displacement of the electrodes from the original train- ing position, the patient’s fatigue, body structure, concentration, motivation and training level.