Real-Time Voluntary Motion Prediction and Parkinson’s Tremor Reduction Using Deep Neural Networks

Wearable tremor suppression devices (WTSD) have been considered as a viable solution to manage parkinsonian tremor. WTSDs showed their ability improve the quality of life individuals suffering from tremor, by helping them perform activities daily living (ADL). Since has shown be nonstationary, nonlinear, and stochastic in nature, performance models used is affected inability adapt nonlinear behaviour Another drawback that limitation estimate or predict one step ahead, which introduces delay when real time with WTSDs, compromises performance. To address these issues, this work proposes deep neural network model learns correlations nonlinearities voluntary motion, capable multi-step prediction minimal delay. A generalized task user-independent presented. The achieved an average estimation percentage accuracy 99.2%. future motion 10, 20, 50, 100 steps ahead was 97.0%, 94.0%, 91.6%, 89.9%, respectively, low 1.5 ms for ahead. proposed also 93.8% ± 1.5% reduction it tested experimental setup time. improvement 25% over Weighted Fourier Linear Combiner (WFLC), estimator commonly WTSDs.