Lie Detection Based on Acoustic Analysis

Area of Science:

• Forensic Acoustics
• Speech Science
• Psycholinguistics

Background:

• Acoustic lie detection offers covert and remote analysis capabilities.
• Traditional polygraphs rely on physiological measures, but acoustic features are gaining interest.
• Developing an acoustic polygraph using phonetic and acoustic features is a novel approach.

Purpose of the Study:

• To construct an acoustic polygraph using phonetic and acoustic features.
• To evaluate the effectiveness of voice onset time (VOT) in lie detection.
• To explore acoustic features as alternatives to traditional physiological measures.

Main Methods:

• A mock crime experiment was conducted with 62 participants (aged 18-30).
• Participants were divided into innocent and guilty groups.
• 31 deceptive and truthful audio recordings were analyzed for voice onset time (VOT).

Main Results:

• Voice onset time (VOT) demonstrated strong performance in lie detection.
• The area under the curve for VOT showed an average sensitivity and specificity of 0.888 (95% CI: 0.803-0.973).
• Other acoustic features indicated a general trend for lie detection, though with lower reference values.

Conclusions:

• Acoustic features, particularly VOT, can effectively aid in lie detection.
• This research supports the development of acoustic polygraphs.
• Future research will investigate additional acoustic and phonetic features for enhanced lie detection.