Human Behavior Analysis Based on Attention Mechanism and LSTM Neural Network

Aiming at the problem that traditional video surveillance algorithm can not take into account the influence of surrounding conditions and the relationship between human behavior, a human behavior analysis method based on LSTM neural network combined with attention mechanism is proposed. This method focuses on the use of key information in surveillance video and related features of human behavior to achieve visual analysis of human behavior. The information of the moving object is obtained by the core region optical flow method between the frames connected to each other. The feature extraction of useful information in different scenarios is achieved through the attention mechanism. The detected moving object information and the extracted features are input into the LSTM neural network to realize real-time analysis of human behavior. The experimental results show that compared with other human behavior analysis methods, this method is more specific in the study of human behavior correlation and has a higher utilization rate for monitoring the surrounding environment information. The rate of human behavior recognition has been significantly improved.