New Insights into Deceptive Speech Presented by Alessandro De Luca at VoiceID
Our own Alessandro De Luca alongside his colleague Volker Dellwo, will present innovative research at the 2nd International Conference on Voice Identity: Perception, Production, and Computational Approaches, held from August 28th-30th 2024 in Marburg, Germany. Their study, "Speakers are Less Discriminable When Producing Deceptive Speech, but Differences Between Deceptive and Truthful Voice Spaces Remain Unclear," explores the complexities of vocal identity in deceptive speech.
Key Findings
The research investigates how deception affects vocal recognition. Using the CSC Deceptive Speech corpus, they discovered that listeners have a harder time distinguishing between speakers when they are lying. Gaussian Mixture Models trained on truthful speech replicated this difficulty, suggesting an acoustic incongruence between truthful and deceptive speech.
Implications
These findings have significant implications for forensic phonetics, security, and AI, highlighting the need for more advanced features to better understand and recognize deceptive speech.
About VoiceID
The conference, held at the Research Center »Deutscher Sprachatlas« in Marburg, Germany, brings together experts from diverse fields such as Psychology, Biology, Linguistics, Physics, Computer Science, and Engineering. It aimes to foster interdisciplinary dialogue on vocal identity, covering topics like acoustic analysis, voice perception and production, neurocognitive mechanisms, computational models, forensic applications, and the ethics of voice AI.
For more details on the conference and presentations, please visit the official website of the 2nd International Conference on Voice Identity.