Gait Analysis Based Parkinson’s Disease Auxiliary Diagnosis System

Abstract

Parkinson’s disease (PD) is a neurodegenerative disease that often occurs in elderly people. Its symptoms are static tremor and slow movement, which affect the life of the patient seriously. With the development of medical technology, the early diagnosis of PD has attracted widespread attention. Many studies have shown that abnormal gait characteristics are potential bases for judging whether suffering from Parkinson’s disease. If PD can be diagnosed in the early stage, it will benefit the control of the disease and subsequent treatment. However, the diagnosis of PD is a complex task which often relies on the doctor’s experience and subjective evaluation. In this stage, because of the lack of professional knowledge of doctors or errors in subjective judgment, it is easy to misdiagnose and miss the best treatment time. In response to this problem, this paper designs an auxiliary diagnosis system for PD based on abnormal gait, composed of embedded devices, mobile terminals and servers. The embedded device uses the accelerometer to collect the patient’s six-dimensional gait data, then the data are transmitted to the mobile phone via Bluetooth and sent to the server. The server analyzes the data by 1D convolutional neural network model and monitors the abnormality of the patient’s gait. Herein, we proved that the use of 1D convolutional neural network for analysis has better performance with five-fold cross-validation, and its recognition accuracy rate reaches 91.4%.