Human body behavior recognition method based on zero sample learning

Abstract

The invention discloses a human body behavior recognition method based on zero sample learning, which improves the classification performance and accuracy of a trained classifier and promotes the realization of automatic labeling of human body behavior categories. The method comprises the following steps: (1) constructing a knowledge graph based on action classes and action associated objects, anddynamically updating the relationship of the knowledge graph through a graph convolutional network AMGCN based on an attention mechanism so as to better describe the relationship of nodes in the graph; (2) learning a generative adversarial network WGAN-GCC based on gradient penalty and cyclic consistency constraint, so that a learned generator can better generate unknown class features; and (3) combining the graph convolution network and the generative adversarial network into a double-flow deep neural network, so that the trained classifier is more discriminative.

Description

technical field [0001] The invention relates to the technical field of computer vision and pattern recognition, in particular to a human behavior recognition method based on zero-sample learning. Background technique [0002] Human action recognition is an important research topic in the field of machine learning and computer vision, and has been widely used in many research topics, such as human-computer interaction, video surveillance, motion retrieval and sports video analysis. At present, with the rapid development of Internet technology and new social media, as well as the continuous expansion of the application field of human-computer interaction technology, the data in the form of images and videos is increasing at an alarming rate every day, and the complexity of human behavior involved is also increasing. It is improving day by day, and the number of video categories is also increasing. Faced with the explosive growth of massive video data, a very difficult problem...

AI Technical Summary

Problems solved by technology

However, the adjacency matrix constructed by these methods remains unchanged after the initial setting, which makes it unable to adaptively describe the changing relationships of nodes in the graph, resulting in incomplete knowledge transfer.

In addition, the existing zero-shot learning method cannot use the sample features of the unknown class in the training, which makes the trained classifier more biased towards predicting the sample category of the known class

Images