BEGIN:VCALENDAR
CALSCALE:GREGORIAN
VERSION:2.0
METHOD:PUBLISH
PRODID:-//Drupal iCal API//EN
X-WR-TIMEZONE:America/New_York
BEGIN:VTIMEZONE
TZID:America/New_York
BEGIN:DAYLIGHT
TZOFFSETFROM:-0500
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
DTSTART:20070311T020000
TZNAME:EDT
TZOFFSETTO:-0400
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0400
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
DTSTART:20071104T020000
TZNAME:EST
TZOFFSETTO:-0500
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
SEQUENCE:1
X-APPLE-TRAVEL-ADVISORY-BEHAVIOR:AUTOMATIC
UID:198116
DTSTAMP:20250127T110855Z
DTSTART;TZID=America/New_York:20250129T140000
DTEND;TZID=America/New_York:20250129T150000
URL;TYPE=URI:https://www.wpi.edu/news/calendar/events/computer-science-depa
 rtment-phd-research-qualifier-samuel-uche-dui-detection-gait-using-multich
 annel
SUMMARY:Computer Science Department, PhD Research Qualifier Samuel Uche  "D
 UI Detection from Gait using a Multichannel 1DCNN-Attention-BiLSTM Framewo
 rk""
DESCRIPTION:\nSamuel Uche\nPhD Student\nWPI - Computer Science\n\nWednesday
 , January 29, 2025\nTime: 2:00 PM - 4:00 PM\nLocation: Fuller Labs 141\nAd
 visor: Prof. Emmanuel Agu\nReader: Prof. Kyumin Lee\nAbstract:\nAlcohol in
 toxication increases Blood Alcohol Content (BAC) and impairs psychomotor, 
 cognitive and motor functions. Driving under the influence (DUI) of alcoho
 l caused over 30% of motor vehicle traffic fatalities in 2017, with $58 bi
 llion accrued in medical, legal and death bills annually. Current measures
  of detecting alcohol DUI - breathalyzers, blood tests, and transdermal al
 cohol monitors - are invasive. Moreover, they require the purchase of addi
 tional hardware and active user involvement and impractical for continuous
  monitoring. Unobtrusive methods of detecting driver intoxication are desi
 rable to reduce DUI incidents. Gait analysis provides a passive, non-invas
 ive approach to continuous DUI detection, enabling unobtrusive monitoring 
 of gait patterns to identify impairment and enhance road safety. Prior wor
 k primarily explored traditional machine learning classifiers, such as Ran
 dom Forest and J48, and some deep learning approaches, but had limitations
  including utilizing handcrafted features that are prone to errors. The de
 ep learning architectures explored achieved suboptimal performance as they
  did not effectively address key challenges such as class imbalance, and g
 ait variability under different conditions.\nThis work proposes a deep lea
 rning approach for automated, passive detection of alcohol intoxication fr
 om smartphone accelerometer sensor data. The challenge of class imbalance 
 that arises from limited intoxicated gait samples was addressed using subj
 ect-level stratified split and random oversampling of intoxicated samples 
 in the training data to ensure balanced class representation. Raw time ser
 ies sensor data were preprocessed using low-pass filtering to reduce noise
 , followed by segmentation into fixed-size windows. Our novel hybrid multi
 channel 1D-CNN-Attention-BiLSTM (MC-Hybrid) framework combines a 1D convol
 utional neural network (1D-CNN) for feature extraction, an attention mecha
 nism for emphasizing critical temporal patterns, and a bidirectional LSTM 
 (BiLSTM) for sequential modeling. This architecture addresses key challeng
 es such as capturing temporal dependencies, highlighting important feature
 s, and improving classification accuracy despite class imbalance.\n\nIn ri
 gorous evaluation, our MC-Hybrid approach achieved an accuracy of 93%, out
 performing the current state-of-the-art by 9.5%, and all baselines by 9.0%
 , with the self-attention mechanism outperforming other attention mechanis
 ms by 2%.\n
END:VEVENT
END:VCALENDAR