Behavior Recognition Based On Genetic Algorithm Optimization BP Neural Network

With the development of information technology, people's increasing demands for social security, the identification as a necessary condition for security, also more and more people to pay attention. Especially in the emerging biometric technology has become mainstream, non-invasive, long-range identification and other features of biological behavior recognition technology has unparalleled advantages. So, become a hot of social studies. This paper has studied gait recognition technology on three points: detection of human behavior, behavior feature extraction, behavior recognition. And, behavior recognition and behavior feature extraction are the most important.In this paper, the use of the image in the background subtraction method to detect human, extracted image morphological processing of the human body with a hollow and noise part, then the image normalization, canny operator used to extract the contours of the human behavior. We use the modified discrete Hu moments to describe human behavior, in order to more accurately describe the characteristics of human behavior, we selected four Hu moments to represent the behavior characteristics. This paper presents a genetic algorithm optimized BP neural network structure, identification of the human behavior, Weizmann behavior using images in the database for training and behavior recognition. After a number of experiments show, this paper presents a genetic algorithm to optimize BP neural network identified the feasibility of human behavior, and obtain a high recognition rate.