Reproduce “Detecting Domain Shifts in Myoelectric Activations: Challenges and Opportunities in Stream Learning” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yibin Sun, Nick Lim, Guilherme Weigert Cassales, Heitor Murilo Gomes, Bernhard Pfahringer, Albert Bifet, Anany Dwivedi
This research investigates the effectiveness of current data stream concept drift detection methods for identifying domain shifts in myoelectric (EMG) signals in real-time. The findings indicate that existing methods perform poorly due to the high intrinsic variability of EMG signals, highlighting the necessity for specialized detection strategies to enhance the long-term reliability of EMG-based assistive technologies.
Key insights
- Domain shifts in myoelectric activations are demonstrably present and significant, confirmed by distinct data segmentations in KPCA visualizations and quantified by notable changes in Kullback-Leibler divergence.
- Existing general-purpose data stream concept drift detection methods are largely insufficient for reliable real-time EMG domain shift detection, yielding overall poor F1 scores (mostly below 0.05, max ~0.4) and high false positive rates.
- The high intrinsic variability of EMG signals *within* a single subject's recording period often exceeds variations *across* subjects or different time periods, causing statistical drift detectors to register benign internal fluctuations as 'drifts' and resulting in an overwhelmingly high false positive rate (>90%).
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.