Research On Lower Limb Motion Intention Recognition Based On SEMG Signal

Exoskeleton robot technology has been widely used in military,life,medical and other aspects.The control of exoskeleton is based on human motion intentions.At present,there are various ways to obtain the motion intention signal of the human body.According to different ways of acquiring information,it can be divided into a motion sensor based method,an image based method and a bioelectric signal based method.By comparing the advantages and disadvantages of various methods,The thesis finally selects the surface EMG signal to identify the human body's intentions.The sEMG signal can be used to accurately,real-time and non-injury to identify the movement intention of the human lower limb before the muscle contraction.The lower limb movement intention can be expressed by the gait phase,and one gait cycle can be divided into multiple phase divisions.The physiological state of the muscles of the lower limbs corresponding to the phase of the asynchronous state is different,so the sEMG signal characteristics of the lower limbs can be extracted to identify the motion intention of the lower limbs.The thesis takes experimental results of the lower limb sEMG signal as the research object,and studys the denoising preprocessing,feature extraction and gait recognition method based on sEMG signal.The thesis proposes an improved SVMKNN algorithm to achieve effective recognition of the lower limb gait phase pattern.The specific research contents are as follows:Firstly,the thesis studies the characteristics of human lower limb movement and divides the gait cycle into five phases: supporting early,middle and late phases,and swing early and late phases.The biceps femoris,tensor fascia late,rectus femoris and gastrocnemius muscles were selected as the source of myoelectric signals by experiments.For the unsteady and susceptible characteristics of myoelectric signals,wavelet denoising and Butterworth filtering were used for pretreatment to improve the signal to noise ratio.Secondly,the thesis proposes a method for detecting the starting point of lower extremity motion and a method for real-time collecting feature vector of sEMG signal through mobile data window technology.which can improve the real-time and accuracy of gait phase recognition.Using time domain and frequency domain analysis methods,various eigenvalues are extracted from sEMG signals,and principal components analysis method is used to extract the features that are most effective for recognition,thereby reducing computational complexity and removing noise,improving recognition efficiency.Then,the SVM algorithm is inaccurately classified near the classification hyperplane,and the KNN algorithm is slow in recognition when the data volume is large.An improved SVM-KNN algorithm is proposed,which calculates the test sample and the optimal hyperplane in the classification stage.IF the distance is more than the given threshold,using the SVM algorithm of the optimized parameter to identify test sample;if the distance is less than the given threshold,the KNN is used to classify all the support vectors as the neighbor samples of the test sample,using the weighted heavy Mahalanobis distance.Calculate the distance between two samples.Finally,the experimental results show that the improved SVM-KNN algorithm can effectively improve the gait phase recognition accuracy and reduce the recognition time.