Effective communication with multiple conversational partners in cocktail party conditions can be attributed to successful auditory scene analysis. Talkers unconsciously adjust to adverse settings by introducing both verbal and non-verbal strategies, such as the Lombard effect. The Lombard effect has traditionally been defined as an increase in vocal intensity as a response to noise, with the purpose of increasing self-monitoring for the talker and intelli-gibility for conversational partners. To assess how the Lombard effect is utilized in multimodal communication, speech and gaze data were collected from four multi-talker groups with pre-established relationships. Each group had casual conversations in both quiet settings and scenarios with external babble noise. Results show that fifteen out of sixteen talkers exhibited an average increase in loudness during interruptive speech in all conditions with-and without external babble noise when compared to unchallenged sections of speech. Comparing gaze behavior during periods of a talkers own speech to periods of silence showed that the majority of talkers had more active gaze when speaking.

Authors: Dourado, Mark; Hassager, Henrik Gert; Udesen, Jesper; Serafin, Stefania
Affiliations: GN Audio Research, Ballerup, Denmark; Aalborg University, Copenhagen, Denmark(See document for exact affiliation information.)
AES Conference: 2022 AES International Conference on Audio for Virtual and Augmented Reality (August 2022)
Paper Number: 22
Publication Date: August 15, 2022 Import into BibTeX
Subject: Paper
Permalink: https://www.aes.org/e-lib/browse.cfm?elib=21852

Download Now (7.7 MB)

This paper is Open Access which means you can download it for free.

Learn more about the AES E-Library

E-Library Location:   /conf/2022/avar/2022_avar_paper_22.pdf

Start a discussion about this paper!