Research on gait recognition and step rate detection based on SVM and auto-correlation analysis

Abstract

In order to monitor the motion state of human body, wearable system is studied. The system can identify gait and detect step rate. By data acquisition system with acceleration sensor, acceleration data are collected under various moving states, and immediately transmitted via blue-tooth. After signals have been analyzed by pre-processing and relevant feature extraction, gait patterns are recognized through support vector machine (SVM) classifier. Compared with back propagation neural network (BPNN) classifier, results indicate that SVM classifier has advantages of simple design and high accuracy for the same data set. So SVM classifier is more suitable. According to crest-detection, auto-correlation analysis algorithm is proposed and tested. Research findings show that auto-correlation analysis can greatly improve the accuracy and adaptability of step rate measurement.