Human action recognition is one of the hotspots in the field of computer vision and pattern recognition. As an important branch of human action recognition, abnormal human action detection has arrested attention of academic and business communities constantly. The research on human action recognition has developed from the research based on human shape features to the research based on gradient design. At present, with the new development of neural network, deep learning has been widely used in action recognition. Because infrared wavebands have advantages of dealing with weak light environment and 24-hour monitoring, they have been applied to the research on human action recognition. This will become a new research hotspot in the field of human action recognition.