Study Of Gait Pattern Recognition Based On Fusion Of Mechanomyography And Attitude Angle Signal

With the increase in the number of disabled people with lower limbs and the gradual aging of the country,it is very important to protect the motor functions of lower limb.First of all,it is significant to achieve the accurate recognition of the lower limbs,which is beneficial for rehabilitation and treatment of disabled patients,detection and monitoring of daily life,as well as the interaction between people and machine,like the application of daily life monitoring of users,and intelligent prosthetics.In the paper,it is the first time to combine the Mechanomyography(MMG)and the attitude angle signal to identify 11 kinds of gait movements(three static motions:stand,sit,squat;four dynamic transition motions:stand to sit,sit to stand,stand to squat,squat to stand;and four dynamic motions:walk,downstairs,upstairs,run).A wireless acquisition equipment was utilized to collect the MMG and the attitude angle of four thigh muscles.After preprocessing and motion segmentation of the MMG signal,the features in the time domain,frequency domain,and time-frequency domain are extracted.Random forest(RF)was used to analyze feature importance,and the relationship between number of features and recognition accuracy based on RF and support vector machine(SVM)was analyzed.Moreover,principal component analysis(PCA)was used for dimension reduction,and the relationship between the reduced dimension and recognition accuracy based on RF and SVM was analyzed as well.Besides,hidden Markov model(HMM)was used for the first time in the pattern recognition based on MMG,and compared with SVM and quadratic discriminant analysis(QDA).In addition,the study was also conducted on the aspects of feature selections,channel combinations,and muscle contribution rates.The results show that the recognition accuracy of dynamic motions based on MMG is 98.27%;the recognition accuracies of static motion and dynamic transition motion based on attitude angle signal are 98.33%and 100%respectively.And the average recognition accuracy of 11 motions is 98.91%.