EMG-Based Essential Tremor Detection Using PSD Features With Recurrent Feedforward Back Propogation Neural Network

Essential tremors (ET) are slow progressive neurological disorder that reduces muscular movements and involuntary muscular contractions. The further complications of ET may lead to Parkinson’s disease and therefore it is very crucial to identify at the early onset. This research study deals with the identification of the presence of ET from the EMG of the patient by using power spectral density (PSD) features. Several PSD estimation methods such as Welch, Yule Walker, covariance, modified covariance, Eigen Vector based on Eigen value and MUSIC, and Thompson Multitaper are employed and are then classified using a recurrent feedback Elman neural network (RFBEN). It is observed from the experimental results that the MUSIC method of estimating the PSD of the EMG along with RFBEN classifier yields a classification accuracy of 99.81%. It can be concluded that the proposed approach demonstrates the possibility of developing automated computer aided diagnostic tool for early detection of Essential tremors.