|
Abstract Human behavior recognition, as an important research hotspot in the field of computer vision, has received much attention in recent years and has broad application prospects in various fields such as smart healthcare, smart agriculture, monitoring and security. With the development of deep learning technology, its application in human behavior recognition has become a trend. Compared to traditional human behavior recognition methods, deep learning based human behavior recognition methods has advantages such as strong adaptability high robustness, fast speed, and higher accuracy, resulting in better recognition results. The article provides a brief introduction to the human behavior recognition method based on manually extracted features, and then sorts out and analyzes deep learning based methods from the perspective of network structure, with a focuse on deep learning network frameworks such as 3D convolutional neural network, dual flow network, recurrent neural network, Transformer, etc. The commonly used datasets in this field, performance of some algorithms on datasets UCF-101 and HMDB-51, were also analyzed and compared. Finally, the current research in this field were summarized and prospected.
|
|
|
|
|
|
2024, Vol. 37 
