Gait recognition method based on modal fusion

Abstract

The invention provides a gait recognition method based on modal fusion. Belonging to the technical field of pattern recognition, in order to solve the problems that an existing gait recognition technology is greatly influenced by illumination conditions and complex backgrounds or covariables, therefore, the contour image extraction is inaccurate. The feature expression capability of the human bodyis limited due to lack of basic attribute information of the human body; meanwhile, because a key point sequence is unstructured data, modeling is difficult to perform by using a deep network, a method of artificially designing features by experience is mostly adopted, effective discrimination features cannot be mined to the maximum extent, and the identification accuracy and the generalization ability are influenced are solved. According to the method, modal fusion is carried out on human body key point features and gait contour map features with complementarity by adopting a modal fusion mode. The human body key point sequence is input into the image convolutional neural network, the gait features based on the skeleton information are extracted, the gait contour image sequence is inputinto the convolutional neural network, the gait features based on the contour information are extracted, and the gait recognition accuracy is improved.

Description

technical field [0001] The present invention relates to the technical field of pattern recognition, in particular to a gait recognition method based on pattern fusion. Background technique [0002] Gait recognition is a very new research direction in the field of computer vision and pattern recognition. It is a new biological feature recognition technology, which aims to realize the recognition of personal identity or the detection of physiological, pathological and psychological characteristics according to people's walking posture. , as a biometric technology, gait recognition has unique advantages that other biometric authentication technologies do not have, that is, it has great recognition potential in the case of long-distance or low video quality, and the gait is difficult to hide or camouflage, etc., but it needs to be designed And it is very complicated and difficult to realize a practical gait recognition system. Because people's walking posture is affected by var...

AI Technical Summary

Problems solved by technology

[0003] 1. Gait recognition through gait contour map or gait energy map is greatly affected by lighting conditions, complex background and covariates, which often leads to inaccurate contour map extraction, thus affecting recognition accuracy

[0004] 2. Carry out gait recognition through skeleton information, and obtain human key point sequences from gait video sequences, but this expression method is limited due to the lack of basic attribute information (such as body shape) of the human body; in addition, due to the key Point sequences are unstructured data, and it is difficult to model them with deep networks. Therefore, the existing technology mostly adopts the method of manually designing features based on experience, so it is impossible to maximize the mining of effective discriminant features, which greatly affects its recognition accuracy and generalization. ability

Images