Abstract
Finger movement recognition based on muscle signals has gained increasing attention in recent years. Machine learning approaches have enabled accurate, non-invasive decoding of voluntary motor activity. Recent advances in deep learning have improved signal interpretation, generalization, and classification performance. However, finger-level applications remain underexplored, particularly in regressionbased estimation and real-time deployment. This study presents a systematic analysis of research trends regarding Machine Learning in Electromyography Finger Recognition over the past decade from 2014 until 2024. A total of 968 documents were collected using a structured search strategy and analyzed using bibliometric tools. The results highlight major developments in signal classification, neural network architecture, and wearable control systems. Co-occurrence and thematic mapping identify emerging research clusters and conceptual structures. Research themes increasingly reflect practical challenges, including signal variability, device integration, and system interpretability. Key gaps were found benchmarking protocols, cross-subject generalizability, and real-time implementation. The limitations restrict translation to assistive, clinical, or commercial domains. This review supports future development of robust, adaptive, and low-latency systems for finger movement recognition. 20142019 focus on feature extraction, classification, and surface EMG signals. 2020-2022 emphasis shifted toward deep learning, CNN, and gesture recognition. 2023-2024 research began targeting real-time control, wearable deployment, and neuromorphic models. Through convergence of machine learning and physiological modeling, EMG systems may soon achieve intuitive and practical finger movement recognition. Such systems will play a critical role in shaping future humanmachine interfaces, prosthetic devices, and neuroadaptive technologies.

This publication has 27 references indexed in Scilit: