https://chahh9808.github.io/tags/exg/ExGHugo Blox Builder (https://hugoblox.com)en-usThu, 15 Jan 2026 00:00:00 +0000https://chahh9808.github.io/media/icon_hu68170e94a17a2a43d6dcb45cf0e8e589_3079_512x512_fill_lanczos_center_3.pnghttps://chahh9808.github.io/tags/exg/https://chahh9808.github.io/post/beyond-hearing/Thu, 15 Jan 2026 00:00:00 +0000https://chahh9808.github.io/post/beyond-hearing/<h2 id="position">Position</h2> <p>Project member in Microsoft Research Asia</p> <h2 id="project-goals--works">Project Goals &amp; Works</h2> <ol> <li>Built a task-agnostic ExG representation learning pipeline for real-world, free-living sensing settings.</li> <li>Proposed Physiology-informed Multi-band Tokenization (PiMT) to represent ExG signals with physiology-aware token decomposition.</li> <li>Trained models with reconstruction-based objectives to capture robust, transferable features across tasks.</li> <li>Evaluated on DailySense and multiple public ExG benchmarks to validate cross-task generalization.</li> </ol> <h2 id="key-results">Key Results</h2> <ol> <li>Demonstrated consistent performance gains over prior ExG baselines across multiple sensing tasks.</li> <li>Showed strong robustness in unconstrained daily-life data collection settings.</li> <li>Highlighted scalability of earphone-based ExG sensing for practical on-device intelligence.</li> </ol> <h2 id="links">Links</h2> <ul> <li>Website: <a href="https://miil.kaist.ac.kr/projects/beyondhearing">https://miil.kaist.ac.kr/projects/beyondhearing</a></li> <li>OpenReview: <a href="https://openreview.net/pdf?id=s79tJrxDmt">https://openreview.net/pdf?id=s79tJrxDmt</a></li> <li>arXiv: <a href="https://arxiv.org/abs/2510.20853">https://arxiv.org/abs/2510.20853</a></li> </ul>https://chahh9808.github.io/publication/beyond-hearing/Wed, 22 Oct 2025 00:00:00 +0000https://chahh9808.github.io/publication/beyond-hearing/