A Parkinson's Disease Detection Method Based On Gait Pressure Signal

In recent years,as China is accelerating its pace into an aging society,the number of patients of Parkinson's Disease is also increasing rapidly,bringing a heavy burden to the patients' families and the society.Traditional clinical diagnosis of Parkinson's disease relies heavily on doctor's subjective opinion so there is a certain chance of misjudgment.Gait analysis is widely applied in the fields of rehabilitation assessment,human identification,and disease diagnosis by systematically studying information of kinematics,dynamics,spatial-temporal parameters,dynamic electromyography,and energy utilization during human walking.Therefore,this paper proposes a method based on gait analysis to identify Parkinson's disease,to provide an objective way to assist doctors in disease diagnosis.The main work of this paper is detailed in the following.1)Pressure data collection and feature extraction:we first collect the original pressure signal of PD patients and controls by means of a platform called U-shape electronical walkway.Then,18 gait parameters are extracted from the original pressure signal,after which a nondimensionalization method is adopted to eliminate the influence of human height to gait features.2)Classification model construction:we build our classification model based on support vector machine algorithm with PSO optimization.Default parameter SVM and grid search SVM are also conducted to compare with PSO-SVM.Extensive experiments show that PSO-SVM outperforms the other two.Experimental results show that the final classification model has an accuracy of 95.66%,a precision of 97.02%,which shows the effectiveness of our proposed method in detecting Parkinson's Disease.