Research On The Method Of Parkinson’s Disease Tremor Evaluation Based On Multi-point Detection

Parkinson’s disease(PD)is a common neurodegenerative disease and it is ranked as the second most common disease in the world after Alzheimer’s disease.Moreover,the number of the people suffering from it has increased drastically,which attracts researchers’ attention.It is normal to judge the severity of this disease from the differences in patients’ symptoms.However,patients are usually not sensitive enough to their disease changes,so these judgment methods,such as scoring tables,will be affected by their own memory.Therefore,it is exceedingly necessary to objectively evaluate the severity of Parkinson’s disease from external assessment.At present,in most cases,judging the severity of Parkinson’s disease often use surface electromechanical images and wearable sensors to collect unilateral tremor signals in PD patients,extract feature parameters,and then analyze the correlation between feature parameters and disease changes,finally select specific parameters to quantify the severity.The disadvantage of this type of method is that the consistency between the selected quantitative parameters and the tremor signal cannot be ensured.Also,single point detection usually selects the most severe side of tremor in PD patients as the signal acquisition point,which will cause the possibility of exaggerating the severity of the disease.In view of the deficiencies of the above research methods,the study on the evaluation method of Parkinson’s disease tremor based on multiple measuring points is proposed.The difference from the quantification method based on single-point detection is that the signal acquisition point lies in the two sides of the PD patient’s body and selects features related to the severity so as to learn the classification model.Compared with single-point detection,this kind of methods is closer to the real disease condition of PD patients and improves the situation of consistency problem between quantitative parameters and tremor signals.The main steps are: using non-contact wearable devices to acquire tremor signals from PD patients,using bandpass filtering to preserve PD tremor in-band signals,and then extracting tremor signal feature parameters according to time and frequency domain analysis to refine the severity of the different conditions To characterize the tremor signal characterization,we finally designed a seismograph based on multi-point seismic seismometers to classify and assess the severity of the disease is finally designed.The main work of this paper is as follows:(1)Make an in-depth study of the grading method of the severity of Parkinson’s disease.Since the current PD tremor assessment method only divides it into "mild","moderate" and "severe" in a vague way,the paper attaches great emphasis on the analysis of the specific classification credentials of UPDRS and Hoehn-Yahr rating scale,as well as their advantages and disadvantages.Finally,it takes UPDRS which covers a wider range of symptoms as classification reference.Thus,in the paper,there are five categories,which are more specific,and a systematic PD tremor assessment method is proposed.(2)Study the noise and characteristics of PD tremor signals.First,use band-pass filters are used to filter noise out of band signals;then,the characteristic quantities of the discrete signal waveforms from the angles of time-domain and frequency-domain;finally,according to the correlation between the feature amount and the tremor assessment,select the main frequency and tremor peaks as the characteristic parameters of the classification.(3)Manage a further study of the classification algorithm,support vector machine-based PD tremor classification model and GBDT-based tremor classification model are designed.And compare the accuracy rate and current rate of classification models based on linear kernel function,polynomial kernel function and radial basis kernel function support vector machine algorithm,and the correct rate and accuracy of the classification model and GBDT classification model in assessing the severity of PD tremors.The tabu search algorithm was used to optimize the parameters of the SVM,which improved the accuracy and accuracy of the classification algorithm.Through the information fusion of the tremor classification results of the bilateral limbs,the decision output obtained is higher than the evaluation results of the unilateral limb.The PD tremor evaluation model based on TS-SVM and DS evidence theory is suitable for the assessment of PD tremor severity.(4)Analysis of the necessity and urgency of establishing a PD tremor assessment system.Based on the obtained classifier,a Parkinson’s disease tremor evaluation system based on multi-point detection was designed and implemented.The environment required for system development is analyzed,and the evaluation algorithm is transplanted into the evaluation system.The Parkinson’s disease tremor evaluation at multiple points is achieved,and the comparison between single-point detection and multi-point detection data and the patient’s disease data analysis of PD are provided.By comparing the accuracy of the PD tremor evaluation method based on the SVM based on three different kernel functions and the PDDT tremor evaluation method based on the GBDT and the complexity of the learning classification model,the SVM tremor evaluation method based on the polynomial kernel after the optimization parameter was found to be classified The accuracy rate and accuracy rate are both higher.Information fusion is performed on the tremor classification results of bilateral limbs,and the resulting decision output is higher than that of unilateral limb evaluation.The PD tremor evaluation model based on TS-SVM and DS evidence theory is also available.Applies to PD tremor severity assessment.