EMG-based Assessments for Rehabilitation Application

University of Technology Sydney. Faculty of Engineering and Information Technology.Biomedical signals–based human control systems have been studied in the biomedical field to improve quality of life. The muscle signal—electromyography (EMG)—is one of the main types of biomedical signals. The muscles are controlled by the central nervous system (CNS). The CNS does not directly control the activation of a large number of muscles, but it still shapes voluntary synergy motion. This research investigated and developed pattern-recognition approaches for EMG signals by studying the automatic body response and voluntary actions to support and understand how the CNS shapes voluntary synergy motion. The purpose of this study is extended to investigate the possible recovery improvement of human rehabilitation movement for stroke patients. This thesis answer core questions: How the automatic body response and voluntary movements can help to improve the quality of life for people with disability?, How can we predict rehabilitation for post-stroke patients?, and how to predict the possible recovery performance ahead of three months?. My doctoral study contributes to knowledge both theoretically and practically. The main research objective was to develop computational intelligence-based EMG for upper limb rehabilitation applications. After building stable procedures for signal processing, we predicted the functional motor recovery of severe, moderate, and mild post-stroke patients during their rehabilitation programs based on support vector machine regression (SVMR). The EMG signals from the upper limb muscles of the patients during their initial rehabilitation sessions were used to train the model. In this thesis we achieved good results with error < 0.5. We developed the non-negative matrix factorisation (NMF) method to extract the synergy EMG to express some features that could support the CNS in shaping the voluntary movement and reducing error. After building our model and extracting the synergy, we calculated the Variance Accounted for Threshold (VAF) to identify the minimum number of synergies that adequately reconstructed the characteristics of the recorded EMGs; our result was > 95% VAF overall. We developed the multilevel mixed-effects (MME) model to predict human recovery based on biomarker assessment sets. We also predicted future rehabilitation for post-stroke patients three months ahead using time series prediction based on synergy EMG. In summary, this pilot study’s results promise the ability to predict the future muscle performance of post-stroke patients based on their current motor ability as well as this summary aims to be easiest for the reader to know upfront everything in the coming chapters