Pattern Recognition Of Head Movement Based On Mechanomyography And Its Application

Mechanomyography(MMG)signal is a low-frequency vibration signal generated by movement of muscle fibers,reflecting the mechanical properties of the muscle.It is a hot spot of research in recent years.Head movement recognition is of great importance in human-machine interaction for convenience of human life,such as flexible control of electric wheelchair.In this study,six head movement patterns,i.e.?bowing head,raising head,side-bending to left,side-bending to right,turning to left,and turning to right have been distinguished by four-channel MMG signals of neck muscles.On the basis of successful identification,MMG signal of the head movement was used to control the hardware devices such as the model car.In this paper,a system was designed to acquire MMG and the system included TD-3 piezoelectric acceleration sensors,and the NI9205 acquisition card,which converted MMG into digital signals to be stored in the computer for subsequent off-line signal processing.Firstly,the original signal was preprocessed by software.Then the effective feature space was extracted according to the characteristics of MMG signals generated by head movement,and the dimensionality of feature space was reduced.Next,the pattern classification results of three classifiers including linear classifier,SVM(Support Vector Machine)and BP neural network were compared and analyzed.The result showed that the optimal features were the feature of bispectral main diagonal slices and energy characteristics of 4 layers decomposition coefficients of coif4 wavelet packets.In dealing with feature reduction,it was found that FLDA had better effect on head movement pattern classification between two dimensionality reduction methods of FLDA and PCA.BP neural network based on cascade forward neural network(CFN)classifier was the best choice.Its recognition accuracy was 91.63%,and the running efficiency was between 2s-4s.Finally,the user graphic interface design of head movement's MMG signal processing based on Matlab GUI was further designed.In addition,the MMG signal of the head movement which was successfully identified was used to control a model car to move in different states,i.e.,moving forward,moving backward,turning left,turning right,accelerating and decelerating,which has great practical significance.