%0 Journal Article
%A CHEN Feng
%A CHEN Xiao-jiang
%A HE Juan
%A XIA Qi-shou
%A YIN Xiao-ling
%T In-Depth Recognition of Human Motion States Based on Smart Phone Perception
%D 2019
%R 10.13190/j.jbupt.2018-221
%J Journal of Beijing University of Posts and Telecommunications
%P 43-50
%V 42
%N 3
%X In order to improve the accuracy of recognition of human motion states by smart phones, an in-depth recognition method based on parallel convolution neural network (PCNN) is proposed. Firstly, the sensor data input format is standardized by using 3D data matrix. Secondly, two PCNNs are used to carry out convolution and pool operation to the acceleration sensor and gyroscope data of human body motion respectively, realizing partial weight sharing. Finally, the two PCNNs are merged in the full-connected layer, and the softmax function is used to classify the human motion states. Experiments show that this method can extract the deep features of human motion states from the original data of the sensor, which can improve the recognition rate of the motion state by comparing with the traditional machine learning method.
%U https://journal.bupt.edu.cn/EN/10.13190/j.jbupt.2018-221