Suffoletto B, Anwar A, Glaister S, Sejdic E. Detection of Alcohol Intoxication Using Voice Features: A Controlled Laboratory Study. J Stud Alcohol Drugs 2023;84(6): 808-813.
This proof-of-concept study explored whether frequency-based analysis of short audio recordings could correctly determine if individuals were intoxicated or not. In this study, twenty participants were assessed in a controlled laboratory setting. Once in the lab, participants completed baseline assessments including measures of disordered alcohol habits (Alcohol Use Disorders Identification Test: AUDIT) and recorded an initial tongue twister (e.g. she sells seashells by the seashore). Participants then consumed a controlled amount of alcohol designed to produce high levels of intoxication and were tasked with reading randomly selected tongue twisters every half hour for up to seven hours. All recordings were done on a smartphone 1-2 feet away from the participant to mirror lifelike distance when not speaking on the phone. During these same checkpoints, individuals’ breath alcohol concentration (BrAC) was also measured. The model developed by the research team was able to accurately detect intoxication (BrAC > 0.08) from changes in soundwave frequency compared to baseline with 97.5% accuracy (95% CI [96.8, 98.2]). This is a promising step in using passive audio data collected by smartphones as a marker of intoxication. Further research is needed to determine if the use of voice as a biomarker for acute alcohol intoxication is viable in naturalistic settings.