Improved Slow Feature Analysis Algorithm and Its Application in Abnormal Human Behavior Recognition

With the development of social economy and the increasing population, human activities are becoming more and more complex, and various abnormal events have also increased. Therefore, abnormal human behavior recognition has gradually become more popular. Because of the complexity of human movement and the diversity of the external environment, it is difficult for us to obtain effective features. In order to solve the problem, an algorithm based on improved slow feature analysis for abnormal human behavior is present. Firstly, the video is preprocessed using the three-frame difference method to obtain human motion foreground. Then, the discriminant slow feature analysis algorithm is adopted to extract the slow features that can represent the nature of the behaviors. The slow features have the invariance of translation, rotation, zoom, illumination, etc., and have the characteristics of direction selectivity and edge direction selectivity. Finally, the cross-validation is used for classification and recognition. The experiments are performed on the KTH public database and the results demonstrate that the proposed algorithm can effectively identify various human behaviors, with an average recognition rate of 92.16%.