Gesture Classification Algorithm Based on SVDD-EPF

Abstract

In the conventional gesture recognition algorithm such as Support Vector Machine (SVM) and Dynamic Time Warping (DTW), the calculation time and accuracy remain to be improved when they face a large number of data to be processed. Therefore, a new algorithm based on Support Vector Domain Description (SVDD) acting on functions is proposed to improve gesture identification. This method mainly uses effective interception data to replenish the length of different gesture data to the same, and generates a continuous and smooth function. On this basis, the SVDD algorithm acts on the extreme points of the function, so the method is simply referred as SVDD-EPF. The results show that the gesture recognition rate of the SVDD-EPF algorithm is higher than that of the traditional SVM algorithm, that is, a better recognition state can be achieved. If it is applied to a mobile phone unlock, it is expected to become a new generation of unlock mode.