Predicting the occurrence of wrist tremor based on electromyography using a hidden Markov model and entropy based learning algorithm. (March 2020)

• Record Type: Journal Article
• Authors: Samaee, Saeedeh; Kobravi, Hamid Reza
• Is Part Of: Biomedical signal processing and control. Volume 57(2020)
• Publication Date: 2020-03
• Subjects: Pathological tremor -- Hidden Markov model -- Maximum entropy model -- Viterbi algorithm; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2019.101739
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: This research specifically is focused on prediction of the occurrence of the tremor. This research is focused on tremors elicited from different disease. The electromyography alongside the Hidden Markov Model used for predicting the occurrence of wrist tremor. Abstract: Pathological tremor is a well-known movement disorder ensues from some diseases such as Parkinson and essential tremor. Develop technologies for tremor suppression is an attractive and open research problem. Incorporating the processing methodologies applicable to the prediction of the occurrence of the tremor burst can corroborate the efficacy of such technologies. Therefore, in this study, a predictive model has been proposed to predict the incidence time of tremor bursts. In the proposed approach, the Markov nonlinear hidden model was employed. The mentioned model was trained once by using the algorithm of Baum Welch and again by combining this algorithm with the maximum entropy algorithm. The Hidden Markov models (HMM) were once trained with raw EMG (Electromyogram) data and by using the extracted features from the EMG signal. The output of the model predicts the occurrence or absence of tremors. The EMG signals were recorded from 11 patients with different pathologic abnormalities. The features such as integrated EMG, mean frequency, and peak frequency were extracted from EMG data and ranked using the RELIEF algorithm. The results showed that the HMM trained with the entropy-based learning method, in the conditions where the EMG signal was its inputs, has the highest performance.