Human Behavior Recognition and Detection Technology Based on Video Stream
Wei Ma, Xian Zhu, Weitao Xiang, Weiwei Li, Zhiguo Wang · 2023
Nowadays, human behavior recognition is a hot spot in computer vision research. As an influential branch of computer vision, human behavior recognition and detection techniques have been widely used in intelligent security, medical monitoring, traffic management, and other fields. The goal of human behavior recognition is to automatically analyze ongoing behaviors from unknown video or image sequences. In this paper, behavior recognition is defined as a fully supervised learning problem, and a multi-task deep neural network is proposed to solve it. Firstly, a 3d convolution network is used to obtain local spatiotemporal context features, and then a recursive convolution neural network is used to input local spatiotemporal context features. Human behavior types can be simultaneously calculated by multi-task neural networks based on global spatiotemporal context.