XAI-Driven sEMG Feature Analysis for Hand Gestures

Surface electromyography (sEMG) signals play a vital role in hand gesture recognition for identifying various hand movements. Due to its non-invasive nature and real-Time muscle activation detection capabilities, sEMG is widely employed in hand gesture identification. The utilization of Machine Learning (ML) for real-Time recognition of sEMG-based hand gestures is an emerging and evolving field. Along with this, the ML classifiers are also known as 'black boxes' since they cannot provide discernible insights into how decisions are produced and what factors impact them. Explainable Artificial Intelligence (XAI) is employed to emphasize the importance of specific features in machine learning models. In this study, employed data related to six hand gestures to assess the performance of five different machine learning models. Among these classifiers, the Extra Tree model demonstrated the highest accuracy, achieving an impressive 85.22%. Furthermore, the XAI-LIME model has been employed to investigate these models' feature relevance for all six hand gestures under examination.