Research on Human Motion Recognition Method Based on Hybrid Convolution Neural Network-Hidden Markov Model

This paper proposes a human motion recognition method based on hybrid Convolutional Neural Network-Hidden Markov Model (CNN-HMM). The main idea is to embed the posterior probability of CNN into the observation probability matrix of HMM. In order to verify the effectiveness of this method, a motion library including one standard motion posture and five non-standard motion postures is established. The experimental data are obtained by combining with the wearable motion capture suit Perception Neuron 2.0 (PN 2.0). The accuracy of leg-lifting is identified and verified by data preprocessing and data segmentation. The results show that the proposed hybrid CNN-HMM method can distinguish the non-standard motion posture from the standard motion postures with a high recognition rate of 97%.