Representative UPDRS Features of Single Wearable Sensor for Severity Classification of Parkinson’s Disease

Parkinson’s disease (PD) is a common neurodegenerative disease. So far, there is no cure for this disease, but the right medicine can slow down the progress of the disease. Therefore, early diagnosis of this disease is very important to improve the quality of life of patients with PD. In recent years, wearable devices have been widely used to classify, predict and monitor the condition of patients with PD. Most previous studies extracted some features for classification, but due to the different research activities, the extracted features lack of standards and generality, when the activities change, the previously extracted features are not necessarily effective. In this paper, we differentiate the PD severity and select representative 20 features related to the disease. For this reason, we designed 8 commom activities and collect data of 85 PD patients using inertial wearable sensors off-the-shelf accelerometer, gyroscope sensors. Our best results demonstrate that the classification accuracy of PD severity is 81.37 \(\%\) . Therefore, this can play a role in assisting doctors in diagnosing and adjusting medication in a timely manner. Meanwhile, feature selection reduced the burden of the model and facilitate the later transplantation of lightweight devices.