DUAN Xiaohu, JIANG Gang, LIU Canghai. Study on human gait recognition based on cell phone acceleration sensor[J]. 2021, 40(1): 30-33+38. DOI: 10.13873/J.1000-9787(2021)01-0030-04.
Human gait dataset of Wireless Data Mining Laboratory of Fordham University(WISDM) collected by mobile phone is used.Realizing the recognition of human gait by support vector machine(SVM)
K nearest neighbour(KNN) and random forest algorithm.The time series of acceleration is processed by windowing and mean smoothing.Taking 2 s as window size
the mean
variance and covariance of triaxial acceleration data are extracted as sample features.The six types of samples(walking
jogging
upstairs
downstairs
sitting
standing) are trained and tested by the above algorithm
and the accuracy
AUC analysis and experimental verification are carried out.Compared with other feature extraction methods and deep learning methods
the method of extracting statistics as feature is simpler and the computational complexity of model is smaller.The average accuracy of linear kernel SVM is 81.3 %.