Research On Quantitative Evaluation And Classification Of Parkinson’s Disease Patients Based On Machine Learning And Gait Analysis

Parkinson’s disease(PD)is a neurodegenerative disease.Compared with the gait of normal people,Parkinson’s disease patients show poorer gait asymmetry and stability.The research on gait is helpful to detect the abnormal gait of patients in time,so that Parkinson’s disease patients can be treated in time.At present,the diagnosis of Parkinson’s disease patients mainly relies on general scales,clinical observations and questionnaire survey methods.Although these diagnostic methods are simple and easy to use,they are not objectively quantifiable.Thanks to the development of machine learning and signal processing,more and more gait features and analysis methods that are important reminders of Parkinson’s disease can be mined from gait signals.In view of the important practical application value of intelligent analysis of gait signals,this paper focuses on the study of quantitative assessment and classification methods for Parkinson’s disease patients based on machine learning and gait analysis,and realizes intelligent diagnosis of Parkinson’s disease based on gait signals.(1)A gait feature extraction method based on the Vertical Ground Reaction Force(VGRF)is proposed.Taking the dispersion degree and fluctuation range of the vertical ground reaction force as the breakthrough point,the quantitative index extraction was carried out,and the difference between the normal gait and the abnormal gait was compared and analyzed.Taking the pressure center as the starting point,preliminary analysis of the difference between pathological and non-pathological pressure centers.Taking symmetry as the starting point,the effectiveness of the proposed features for abnormal gait detection is analyzed.(2)A feature extraction method based on nonlinear dynamics is proposed.Gait signal is a kind of dynamic signal,and dynamic change is its essential feature.It is necessary to further extract effective dynamic features on the basis of the above to improve the recognition rate.Starting from the theory of nonlinear dynamics,the power spectrum method is used to analyze the distribution and periodicity of the frequency spectrum to determine the chaotic characteristics of the time-varying gait signal.The complexity of the gait signal is studied from the internal structural changes of the system,Extract LZ and C0 complexity;tudy the entropy value from the perspective of the information change of the dynamic system,and Three kinds of entropy are extracted.Analyze the eigenvalue distribution of PD patients and healthy subjects to determine the validity and separability of gait signal features.(3)Propose the research of Parkinson’s disease classification method based on machine learning.The first step is to realize the classification and recognition of Parkinson’s disease based on a single feature through machine learning algorithms.In the second step,in order to deal with the problem that a single feature is difficult to describe the complex gait movement comprehensively and accurately,and a feature fusion method is proposed.Contrast and analyze cascade feature fusion,PCA feature fusion,KPCA feature fusion.In the third step,in order to deal with the problem that the new feature after feature fusion no longer has the specific meaning contained in the original feature and is of high complexity,a number of excellent features that best reflect the essence of the feature set are selected through the feature selection algorithm to construct a dimensionality reduction space.(4)Developed an automatic analysis and diagnosis system for Parkinson’s disease.Taking symmetry as the starting point,complete the development of gait asymmetry quantification and classification system,it constitutes an automatic analysis and diagnosis system for Parkinson’s disease with multiple features and multiple indicators.This system realizes the import of the gait database and the graphic display of the original VGRF signal.Through feature extraction and feature selection of the VGRF signal,the automatic classification and recognition of Parkinson’s disease is realized.