Open AccessA High-Performance Gait Recognition Method Based on n-Fold Bernoulli Theory
Author(s) -
Qing Zhou,
Jarhinbek Rasol,
Yuelei Xu,
Zhaoxiang Zhang,
Lujuan Hu
Publication year - 2022
Publication title -
ieee access
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.587
H-Index - 127
ISSN - 2169-3536
DOI - 10.1109/access.2022.3212366
Subject(s) - aerospace , bioengineering , communication, networking and broadcast technologies , components, circuits, devices and systems , computing and processing , engineered materials, dielectrics and plasmas , engineering profession , fields, waves and electromagnetics , general topics for engineers , geoscience , nuclear engineering , photonics and electrooptics , power, energy and industry applications , robotics and control systems , signal processing and analysis , transportation
Gait feature recognition refers to recognizing identities by collecting the characteristics of people when they walk. It shows the advantages of noncontact measurement, concealment, and nonimitability, and it also has good application value in monitoring, security, and company management. This paper utilizes Kinect to collect the three-dimensional coordinate data of human bones. Taking the spatial distances between the bone nodes as features, we solve the problem of placement and angle sensitivity of the camera. We design a fast and high-accuracy classifier based on the One-versus-one (OVO) and One-versus-rest (OVR) multiclassification algorithms derived from a support vector machine (SVM), which can realize the identification of persons without data records, and the number of classifiers is greatly reduced by design optimization. In terms of accuracy optimization, a filter based on n-fold Bernoulli theory is proposed to improve the classification accuracy of the multiclassifier. We select 20000 sets of data for fifty volunteers. Experimental results show that the design in this paper can effectively yield improved classification accuracy, which is 99.8%, and reduce the number of originally required classifiers by 91%-95%.