Study On Human Action Recognition Model Based On Relevance Vector Machine

Human action recognition-related research has received high attention in interdisciplinary fields such as statistical learning,computer science,and rehabilitation engineering.How to accurately characterize significant differences in human motion patterns based on small sample human motion data and improve their recognition accuracy is a hot issue in recent years.Therefore,this paper attempts to capture the inherent differential sparse distribution characteristics of human action patterns based on the relevance vector machine Bayesian learning algorithm,in order to improve the performance of classification generalization based on small sample human action data,and offer new research ideas and methods for correlative research.The related work is as follows:(1)A new model for human action discrimination based on correlation vector machine and XGBoost algorithm is proposed.The model makes full use of correlation vector machine Bayesian learning model posterior probability,and defines information entropy to measure the separation of different modes in a small sample human action data space.The sample points can accurately represent the difference distribution of different human action patterns,and effectively improve the accuracy of human action classification of the XGBoost algorithm.The results show that the classification accuracy of the proposed model in this paper can reach 87.2%,which can effectively represent the difference distribution of human action patterns and improve its classification and generalization performance.(2)A new model of human action classification based on kernel principal component analysis is proposed.The model makes full use of the kernel function to integrate kernel principal component analysis and correlation vector machine,and mines more human bodies in the high-dimensional characteristic space.The nonlinear characteristics of action difference information can improve the Bayesian learning algorithm to capture the inherent difference sparse distribution characteristics of human actions,and improve the performance of classification generalization of small sample human action data.The results show that the proposed model can overcome the shortcomings of small sample data,such as poor sparse representation of human action differences,only about 10 correlation vectors are needed to accurately represent the sparse distribution of human action patterns,and the classification accuracy can reach 95.8%,effectively improving.The classification and generalization performance provides new ideas and methods for improving the accuracy of human action recognition.