Behavior Recognition Method Based on Independent Subspace Network

Abstract

The invention discloses a behavior recognition method based on an independent subspace network; it specifically includes the following steps: preprocessing and learning spatial features, extracting spatiotemporal features, representing video as a histogram vector, training an SVM classifier and behavior recognition. The behavior recognition method based on the independent subspace network of the present invention extracts spatial features from video data, and then pools these spatial features to obtain features equivalent to spatio-temporal features, and then uses the bag of words method to convert each video The sequence is represented as a histogram vector of each word in the bag of words, and then use multiple trained SVM classifiers to apply a one-to-many strategy for multi-category behavior recognition, increase the robustness of behavior recognition, and improve behavior recognition Rate.

Description

technical field [0001] The invention belongs to the technical field of artificial intelligence recognition, and in particular relates to a behavior recognition method based on an independent subspace network. Background technique [0002] With the rapid advancement of urbanization in China, the sharp increase in urban population and the rampant international terrorism, urban public safety and personal safety have become the research focus of intelligent monitoring research institutions or related technology research and development departments. Behavior recognition technology that analyzes and recognizes human behavior is the top priority of this research. At present, action recognition is an important hot research problem in the field of artificial intelligence. It is widely used in public safety intelligent monitoring, human-computer interaction, smart home, patient or elderly safety monitoring, etc. The key technology of behavior recognition is the realization of behavi...

AI Technical Summary

Problems solved by technology

This patent uses SURF points of interest to construct the trajectory representing the human body. Although this method uses some denoising methods, there are many noises in the behavioral video, such as the occlusion of the human body by other people or objects in the scene, noisy background, etc. will have a great impact on the method, thereby reducing the recognition effect of the behavior

Moreover, this method uses traditional human-designed features to represent behaviors, and these hand-designed features have great limitations in identifying behaviors.

Images