Multi-perspective gait classification method based on multiple regression

Abstract

The invention discloses a multi-perspective gait classification method based on multiple regression. The method comprises the steps that 1 a human body contour is extracted from a gait data set to be classified; two human body area images of inner point representation and boundary representation are constructed, and a hub energy diagram is further generated; 2 a multi-period hybrid gait energy matrix corresponding to each gait image sequence is calculated as a gait feature based on the hub energy diagram of the previous step; and 3 multi-perspective gait classification is converted into multiple regression, and a convolution neural network is constructed for solving. According to the invention, the multi-perspective gait classification method based on multiple regression uses the multi-period hybrid gait energy matrix as the gait feature; accurate gait period segmentation is eliminated; the reliance on gait period segmentation is reduced; multi-perspective gait classification is converted into multiple regression; the convolution neural network is constructed for solving; and the gait classification accuracy is improved.

Description

technical field [0001] The invention belongs to the technical field of biological feature processing in pattern recognition, in particular to a multi-view gait classification method based on multiple regression. Background technique [0002] Gait classification is one of the most potential biometric processing technologies under long-distance detection conditions. It can realize identity recognition and lower limb health status monitoring based on people's walking posture under conditions of changing clothing and viewing angles. The basic steps of gait classification are: first, the gait image is separated from the gait dataset to be classified by moving object detection. Secondly, detect the gait cycle, segment a series of periodic gait image sequences, and extract gait features from them. The selection and accurate extraction of gait features is one of the most important links in gait classification, which will directly affect the subsequent classification accuracy. Fina...

AI Technical Summary

Problems solved by technology

When extracting gait features, the existing gait classification algorithms need to segment the gait cycle first, which has a serious dependence on factors such as time and pace, which limits the application and promotion of gait classification

[0003] Compared with the processing technology of biological characteristics such as face, fingerprint and iris, due to the relatively short development time of gait processing, the current gait classification and recognition technology is still in the research stage, lacking systematic methods and theories

Images