Research Of Electromyography Pattern Recognition Based On Support Vector Machine

With the development of technology, the intelligent prosthesis has replaced the traditional prosthesis gradually. Its main feature is that it can tune the parameters automatically according to the circumstance. But it can not change according to the frequency of step and angle of the knee. This study proposes a new type of intelligent prosthesis whose main control source is electromyographic signal. It can tune the parameters of prosthesis system automatically according to human's consciousness and circumstance, and work agilely and stably. Since it is easy to be extracted without hurt, surface electromyographic (sEMG) signal is the ideal control source. This study is the basic research in order to realize this goal.Based on the pretreatment and extraction features of lower limb sEMG signal, recognizing different terrains and approaching the angle of knee joint is realized. The main content and innovations are shown as follows:1. Extracting the sEMG features from lower limbThe features extraction is an essential course in model identification. The identification capacity of an identification system directly relates with selection of feature vector. For a prosthesis controlled by sEMG, the ultimate problem of the motional pattern recognition is how to find out the effective characteristic signal. With the view to sEMG signal features, this paper recommend decomposed wavelet packet coefficients and the energies of each frequency band as the feature vector.2. Classification of the sEMG feature vector by support vector machineIn order to identify the different terrain, such as flat, slope and stair, support vector machine is used to classify by energies of each frequency band as the feature vectors. The experimental results indicate that this method can perform very well in classification capability when compared to the neural networks with LVQ applied as classifiers in the algorithm, and has a great potential in practical application of artificial lower limb.3. Predicting the angle of knee joint by support vector regression modelIn order to identify angle of knee joint, support vector regression model is proposed. By the knee joint angle curve got from 3-D Gain Analyzer, we approach the knee joint angle curve to achieve the most approximation between prosthesis and body. Through the combination of support vector regression and fuzzy algorithm, the performance of algorithm is improved.