Ref-EMGBench: Benchmarking Reference Normalization for Electromyography Data
Keywords: EMG, reference normalization, domain adaptation
TL;DR: We present a comprehensive benchmarking of five statistical normalization methods, evaluating their effectiveness in mitigating intersubject variability for hand gesture recognition based on EMG data.
Abstract: Electromyography (EMG)-based hand gesture recognition is essential for applications in prosthetics, rehabilitation, and human-robot interaction. Despite advances in machine learning, domain shift caused by intersubject variability often leads to degraded model performance when applying trained models to new users. In this study, we revisit the statistical reference normalization methods to mitigate the domain shift in EMG data in a leave-one-subject-out train-test split setting. We systematically benchmark five popular amplitude-based normalization techniques to assess their effectiveness in subject-specific classification with varied datasets and percentages for normalization. Experimental results show that Min-Max and Peak normalization outperform others, yielding higher classification accuracy on EMG data. We further visualize the domain shifts in the feature space throughout the training process and provide an analysis based on EMG signal characteristics. Our findings indicate that proper normalization significantly reduces inter-subject variability of EMG samples, enhancing model adaptation and providing insights for bridging domain shifts in future EMG-based gesture recognition research. The benchmark code for domain adaptation approaches on EMG signals is available at ref-emgbench.github.io.
Primary Area: applications to neuroscience & cognitive science
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 13429
Loading