A Comparative Analysis of Different Features for EMG Signal Classification

Abstract

Electromyography (EMG) signals serve as vital tools in neurological and neuromuscular conditions diagnosis. Various features are used as inputs for pattern recognition algorithms. This project intends to increase the precision and efficacy of prosthetic limb control, with the goal of boosting the quality of life for individuals with limb amputations, using a Linear Support Vector Machine technique. Specifically, we intend to analyze the usefulness of the distinctive feature known as Cardinality within diverse combinations of time-domain and frequency-domain features. In order to improve signal quality, the raw EMG signal is filtered and segmented. The time-domain and frequency-domain features are then retrieved from overlapping segments, and the most relevant ones are retained using exhaustive feature selection. An SVM classifier is then used to examine the possible impact of Cardinality on prosthetic control and rehabilitation outcomes. The research findings show that the efficiency of Cardinality is dependent on the precision of the units used. Cardinality performed best when seven decimal points are used. MAV stands out among time-domain features, as it generated a high number of combinations with Cardinality, enhancing its performance in myoelectric pattern recognition and BP emerges as the top-performing frequency-domain feature when integrated with Cardinality, surpassing other frequency-domain features and forming the most numerous combinations. The SVM classifier achieved classification accuracy of 85.58% of M1, 70.49% of M2, 77.32% of M3, 77.24% of M4, 80.82% of M5, 77.52% of M6, 82.94% of M7, 84.34% of M8, 84.75% of M9, 86.92% of M10 for the combination of Cardinality with MAV and BP. As advancements in prosthetics and rehabilitation technologies continue, the insights gained from this study can play a pivotal role in refining the precision and efficiency of Myoelectric Control systems, ultimately benefiting individuals with limb loss or motor impairments.