Detection of Leg Tremors in Parkinson's Disease Patients: An Experimental Wearable Leg Band Solution

Parkinson's disease (PD) is a gradually worsening disorder of the nervous system that affects a person's ability to control their movements. As the disease gets worse, it makes it harder to move normally. Key symptoms include tremors at various stages. The usefulness of tremor characteristics for PD early identification and monitoring is examined in this study. Data from wearable sensors on (n=14) PD patients and normal volunteers, recorded during activities averaging 200±15 minutes, was used. A sliding window and segmentation technique were used to extract time-domain and frequency-domain characteristics. Machine learning models like K-Nearest Neighbour (KNN), Neural Networks (NN), Discriminant Analysis (DA), Decision Trees (DT), Support Vector Machine (SVM), and Naïve Bayes (NB) were employed to identify distinguishing characteristics of Parkinson's disease (PD) patients compared to healthy controls. Validation was conducted using a leave-one-subject-out (LOSO) approach. The Linear Discriminant model had the lowest accuracy at 89.5% and the greatest F1 score of 90.4%, while the Fine K-NN model achieved the highest accuracy of 98.3% and an F1 score of 98.9%. These findings show how these ML techniques may be used to detect tremors in PD patients.