基于机器学习的生物雷达非接触识别头部特定动作方法研究OACSTPCD

Objective To propose a new method of classification of specific head movement based on non-contact bioradar sensor.Methods Firstly,radar signals were collected for specific head movements——static,nodding,left turn,right turn,leaning back,opening mouth,nodding left and nodding right.Secondly,two kinds of image data were obtained by time domain processing and time frequency analysis and the principal component analysis(PCA)was used to construct new integrated features that were more effective.Thirdly,the support vector machine(SVM)model was used to classify the new features.Results Based on PCA and SVM,a human motion feature extraction method was constructed to recognize human head movements.The results showed that the classification recognition accuracy of different types of head movements could reach 88.64% .Conclusion The non-contact recognition of head movements proposed in this paper is of great significance for enhancing social communication,delaying the development of disease and improving the quality of life of Alzheimer's patients.