In this paper we present a study of the influence of alcohol intoxication on an automatic speaker verification system. It is widely known that alcohol intoxication affects one's speech in many ways, but it remains to be studied how well a system can recognise a person affected by alcohol intoxication. Using the Alcohol Language Corpus as speech database and a text-independent GMM-UBM speaker verification system, we perform experiments to analyse the effects of alcohol intoxication in detail. In different experimental setups, using recordings in either sober or alcoholised condition for speaker enrolment or testing, the influence of intoxication on the error rate of the speaker verification system is investigated. Compared to the baseline experiment without alcohol intoxication, the results indicate a generally negative influence of alcohol intoxication on the verification system. This influence is larger for female speakers compared to males.
Geiger, Jürgen; Zhang, Boxin; Schuller, Björn; Rigoll, Gerhard
Affiliations: Institute for Human-Machine Communication, TU München, Munich, Germany; Imperial College London, London, UK(See document for exact affiliation information.)
AES Conference: 53rd International Conference: Semantic Audio (January 2014)
Paper Number: 4-1
Publication Date: January 27, 2014
Subject: Speech Processing and Analysis
No AES members have commented on this paper yet.
If you are not yet an AES member and have something important to say about this paper then we urge you to join the AES today and make your voice heard. You can join online today by clicking here.