Hand Gesture Recognition Based on Semg Signal and Improved SVM Voting Method

To explore how to use as few features as possible to recognize as many gesture types as possible, a surface electromyographic (sEMG) gesture recognition algorithm based on an improved support vector machine (SVM) voting method is proposed. Compared with the traditional voting method, there are two main differences. First, all SVMs can be in different sub-training sets, where each channel corresponds to a sub-training set, and the sub-training set where each SVM is located is determined by a class distance-based separability criterion. Secondly, a separability enhancement method between two types of samples is proposed. Before training, each SVM is subjected to separability enhancement processing. Experimental results show that the dimensionality of the feature vector using the improved SVM voting method is only three-dimensional, and it can recognize twelve gestures with 84% accuracy. The recognition effect is better than several other methods, and the effectiveness is verified.