Abstract:
Surface Electromyographic (sEMG) signals have become prevalent in a variety
of applications, such as human-computer interface, rehabilitation, and medical
diagnostics. To improve the classification of sEMG signals and clarify their link with
various motor neuron activities, this study explores new directions in signal processing
and multi-scale analysis. The study aims to explore and understand the intricate
dynamics of neuromuscular control through electromyographic (EMG) signal
processing and multiscale analysis, elucidating fundamental mechanisms underlying
movement execution and coordination. By analyzing EMG signals using multiscale
analysis, researchers unveil intricate patterns of muscle activation, offering insights into
single motor unit firings and coordinated movements involving multiple muscle groups.
Through meticulous examination, the research unveils the correlation between surface
electromyographic (sEMG) signals and motor neuron functions, highlighting potential
applications in medical diagnostics and rehabilitation robotics.
The study begins with a systematic literature review (SLR) that provides a
comparative overview of recent research on sEMG-based hand gesture recognition
systems, identifying gaps and evaluating data collection, processing, and classification
algorithms. The research presents a simple approach to decomposing sEMG signals,
crucial for understanding muscle activation patterns and improving prosthetics and
ergonomic interfaces. Utilizing Maximal Overlapping Discrete Wavelet Transform
(MODWT) for signal decomposition, the research achieves up to 94% accuracy in
identifying neural activity.
Correlation analysis reveals discriminative features for differentiating signals,
enhancing classification accuracy for finger movements. The insights gleaned from the
correlation analysis pave the way for future investigations into the complexities of
neuromuscular function and motor control mechanisms. A proposed algorithm for
classifying finger gestures demonstrates the effectiveness of processing raw sEMG
signals and extracting dominant features using machine learning classifiers. With an
ii
average classification accuracy of 94.15% from the observed dominating channels, the
study emphasizes the importance of an effective model for myoelectric pattern
recognition systems in controlling prosthetic limbs.
As the research advances, its implications hold promise for enhancing the
precision and efficacy of various neurorehabilitation strategies and augmenting our
understanding of human motor control mechanisms. The benefits and drawbacks of the
algorithms and techniques used to identify, process, and quantify certain patterns and
properties of myoelectric signals were covered. These techniques pave the way for more
effective clinical diagnoses and rehabilitation interventions, advancing our
understanding of human movement and enhancing motor function restoration.
