BEGIN:VCALENDAR VERSION:2.0 PRODID:IEEE vTools.Events//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:America/Los_Angeles BEGIN:DAYLIGHT DTSTART:20260308T030000 TZOFFSETFROM:-0800 TZOFFSETTO:-0700 RRULE:FREQ=YEARLY;BYDAY=2SU;BYMONTH=3 TZNAME:PDT END:DAYLIGHT BEGIN:STANDARD DTSTART:20251102T010000 TZOFFSETFROM:-0700 TZOFFSETTO:-0800 RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=11 TZNAME:PST END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20251230T035358Z UID:C790EBA9-7C53-48A4-916C-CB2B7B979BC8 DTSTART;TZID=America/Los_Angeles:20251222T115000 DTEND;TZID=America/Los_Angeles:20251222T130000 DESCRIPTION:Longitudinal and continuous monitoring of cough is crucial for early and accurate diagnosis of respiratory diseases. Recent developments in wearables provides at-home remote symptom monitoring with respect to mo re accurate and less frequent assessment in the clinics\, but face practic al challenges such as speech privacy\, poor audio quality and background n oise in uncontrolled real-world settings. Our work addresses these challen ges by developing and optimizing a multimodal cough detection system\, enh anced with an Out-of-Distribution (OOD) detection algorithm. The cough sen sing modalities include audio and Inertial Measurement Unit (IMU) signals. The system is optimized through training with an enhanced dataset and a w eighted multi-loss approach for in-distribution classification\, while OOD detection is improved by reconstructing training data components. Experim ents demonstrate robustness across window sizes from 1–5 seconds and eff ectiveness at low audio sampling rates\, where privacy is preserved. The o ptimized system achieves 90.1% accuracy at 16 kHz and 87.3% at 750 Hz\, ev en with half the inference data being OOD. Most misclassifications arise f rom nonverbal sounds (e.g.\, sneezes\, groans). Overall\, the proposed Aud io-IMU multimodal model with OOD detection significantly improves cough de tection performance and offers a practical solution for real-world wearabl e applications. Wearable devices with on-board neural acceleration capabil ities have been developed to enable fusion of air and bone microphones\, a nd inertial measurements together with real-time processing.\n\nSpeaker(s) : Prof. Dr. Edgar J Lobaton\n\nAgenda: \n11:50 am to 12:00 pm: Social\n\n1 2:00 pm to 12:45 pm: Talk\n\n12:45 pm to 1:00 pm: Q/A\n\nVirtual: https:// events.vtools.ieee.org/m/523030 LOCATION:Virtual: https://events.vtools.ieee.org/m/523030 ORGANIZER:swapnilsayansaha@ieee.org SEQUENCE:20 SUMMARY:Robust Multimodal Cough Detection with Optimized Out-of-Distributio n Detection for Wearables URL;VALUE=URI:https://events.vtools.ieee.org/m/523030 X-ALT-DESC:Description: <br /><p>Longitudinal and continuous monitoring of cough is crucial for early and accurate diagnosis of respiratory diseases. Recent developments in wearables provides at-home remote symptom monitori ng with respect to more accurate and less frequent assessment in the clini cs\, but face practical challenges such as speech privacy\, poor audio qua lity and background noise in uncontrolled real-world settings. Our work ad dresses these challenges by developing and optimizing a multimodal cough d etection system\, enhanced with an Out-of-Distribution (OOD) detection alg orithm. The cough sensing modalities include audio and Inertial Measuremen t Unit (IMU) signals. The system is optimized through training with an enh anced dataset and a weighted multi-loss approach for in-distribution class ification\, while OOD detection is improved by reconstructing training dat a components. Experiments demonstrate robustness across window sizes from 1&ndash\;5 seconds and effectiveness at low audio sampling rates\, where p rivacy is preserved. The optimized system achieves 90.1% accuracy at 16 kH z and 87.3% at 750 Hz\, even with half the inference data being OOD. Most misclassifications arise from nonverbal sounds (e.g.\, sneezes\, groans). Overall\, the proposed Audio-IMU multimodal model with OOD detection signi ficantly improves cough detection performance and offers a practical solut ion for real-world wearable applications. Wearable devices with on-board n eural acceleration capabilities have been developed to enable fusion of ai r and bone microphones\, and inertial measurements together with real-time processing.</p><br /><br />Agenda: <br /><p>11:50 am to 12:00 pm: Social< /p>\n<p>12:00 pm to 12:45 pm: Talk</p>\n<p>12:45 pm to 1:00 pm: Q/A</p> END:VEVENT END:VCALENDAR