Agu, Emmanuel O.
Driving under the influence is one of the largest risk factors leading to accidents. Intoxication manifests in the drinker's voice. This paper explores deep learning architectures and hand extracted features to classify voice samples as either intoxicated of sober. Our method classifies intoxicated speech with an unweighted average recall of 59.2%.
Worcester Polytechnic Institute
Major Qualifying Project
All authors have granted to WPI a nonexclusive royalty-free license to distribute copies of the work, subject to other agreements. Copyright is held by the author or authors, with all rights reserved, unless otherwise noted.