Auditory-Visual Speech Processing 2005

British Columbia, Canada
July 24-27, 2005

Facilitating Speech Detection in Style!: The Effect of Visual Speaking Style on the Detection of Speech in Noise

Nicole Lees, Denis K. Burnham

University of Western Sydney, MARCS Auditory Laboratories, Australia

Speakers naturally modify the way they produce speech depending on the listening environment. A hyper-articulatory speaking style is typically employed when listening circumstances are difficult. In contrast, hypo-articulation is adopted when the elements for listening are favourable. Previous research suggests that the intentions of the speaker are realized in listener's recognition of spoken utterances [1]. Given recent findings that the sight of articulating faces increase the detectability of speech in noise [2, 3, 4, 5]; the present study investigates whether speech in noise is more detectable when the listener views hyper-articulated compared to hypo-articulated speech. Normal, Hyper- and Hypo-articulated styles of ``ba, bi, bu, da, ga, tha'' spoken by three speakers were paired with corresponding static images (auditory information only) or dynamic (audio-visual condition) visual articulations in a two interval (noise only/speech plus noise) forced choice detection task at three signal-to-noise ratios (0dB, -2dB, and -4dB). Seeing the articulating face provides a significant advantage for detecting speech in noise compared to auditory only presentation regardless of speaking style. Additionally, hyper-articulated speech offered a significant advantage over hypo-articulated speech suggesting that the amount of facial movement may modulate the AV facilitation effect in the detection of speech in noise.

Full Paper

Bibliographic reference.  Lees, Nicole / Burnham, Denis K. (2005): "Facilitating speech detection in style!: the effect of visual speaking style on the detection of speech in noise", In AVSP-2005, 23-28.