Motion Intent Recognition Of Lower Limb Prosthesis Based On Wavelet Transform

Intelligent lower limb prosthesis in terms of function and safety than ordinary prosthesis has a more obvious improvement,it can reduce the metabolic energy and reacts actively,to help amputees to carry out daily movement,so as to restore the patient’s ability to work and live.Therefore,the intelligent lower limb prosthetic system needs to automatically identify the motion intent of the amputee,adjust the control strategy of the prosthesis system,timely and accurately obtain the motion intent of the person and execute the corresponding movement,so that the movement of the amputee can reach or approach the healthy limb,achieve safe and smooth walking.Motion intent recognition of intelligent lower limb prosthesis can be considered as a class of short-time behavioral recognition,where the main problem is to explore the gait transients between two adjacent different gait patterns.Most intent recognition methods usually collect motion data from the affected side of the subject,extract features such as root mean square,standard deviation,mean value,most value to construct feature vectors,and apply classifiers to recognize motion patterns.The data used for intent recognition are collected using multiple types of sensors and the number of sensors is high,and the dimension of the constructed feature vectors is high,which brings problems such as large computational complexity.The short-time samples obtained by transient transformation have some instability in extracting statistical features,which may lose motion information and reduce the classification accuracy.Based on this,this paper introduces the wavelet transform applied to the motion intent recognition of intelligent lower limb prosthesis.On the one hand,one of the main problems of motion intent recognition is to explore the gait transient transitions between two adjacent different steady-state modes.At the same time,the human lower limb movement has inherent continuity,and it is crucial to explore features that reflect the inherent continuity of movements for intention recognition.Based on the above considerations,the dual-tree complex wavelet transform(DTCWT)is applied to the motion intent recognition of lower limb prosthesis in Chapter 2 of this paper.The local analysis capability of wavelet transform can amplify the transient changes of gait information and facilitate the extraction of transient change information between two adjacent different steady states.The translation invariance and direction selectivity of DTCWT help characterize the continuous features of motion patterns,which better reflects the inherent continuity that human lower limb movements.On the other hand,the key issue of motion intent recognition is to obtain the human motion intent in a timely and accurate manner,and constructing feature vectors with fewer dimensions is very important for intent recognition.In order to solve the problems of high dimension of feature vectors and large computational complexity.In Chapter 3,this paper introduces the double-density dual tree complex wavelet transform(DD-DTCWT)method,which has the advantages of both dual-tree complex wavelet and dual-density wavelet,and the DD-DTCWT is applied to the motion intent recognition of intelligent lower limb.Firstly,DD-DTCWT has translation invariance and direction selectivity,which is conducive to exploring gait transient transition information and can better distinguish the patterns with strong motion similarity.Secondly,considering that the motion gait is directional,the directional selectivity of DD-DTCWT has been further improved compared to that of DTCWT,thus better extracting the instantaneous directional change information of motion.Finally,according to the correlation of wavelet coefficients,the wavelet coefficients are processed by combining the mean and variance,thus the dimensionality of the feature space is reduced.