Human Detection and Recognition Based On Gradient Histogram and Fast Convergence Algorithm

Abstract

Human detection and recognition is a core task for video analysis, and the extraction and selection of behavior features will affect the recognition effect directly. Based on the current human body recognition method combining the histogram of directional gradients and fast convergence algorithms, the features if fast convergence algorithm and histogram of oriented gradients (HOG) are proposed.Based on which, the fusion features are formed, and then the behavior recognition method of features classification is completed by using the support vector machine (SVM). First, the effectiveness of the algorithm is verifies by combining the research of Matlab human behavior recognition and detection and adopting the KTH and Weizmann human behavior library. The experimental results show that the algorithm has improved the recognition efficiency in human behavior recognition, which also has a good recognition effect in human detection and obtains a better recognition rate.