Auditory "bubbles": Efficient classification of the spectrotemporal modulations essential for speech intelligibility

• Published in: The Journal of the Acoustical Society of America Vol. 140; no. 2; p. 1072
• Main Authors: Venezia, Jonathan H, Hickok, Gregory, Richards, Virginia M
• Format: Journal Article
• Language: English
• Published: United States 01.08.2016
• Subjects: Adult; Auditory Threshold; Female; Humans; Male; Psychoacoustics; Speech Intelligibility - physiology; Speech Perception
• ISSN: 1520-8524
• DOI: 10.1121/1.4960544

Speech intelligibility depends on the integrity of spectrotemporal patterns in the signal. The current study is concerned with the speech modulation power spectrum (MPS), which is a two-dimensional representation of energy at different combinations of temporal and spectral (i.e., spectrotemporal) modulation rates. A psychophysical procedure was developed to identify the regions of the MPS that contribute to successful reception of auditory sentences. The procedure, based on the two-dimensional image classification technique known as "bubbles" (Gosselin and Schyns (2001). Vision Res. 41, 2261-2271), involves filtering (i.e., degrading) the speech signal by removing parts of the MPS at random, and relating filter patterns to observer performance (keywords identified) over a number of trials. The result is a classification image (CImg) or "perceptual map" that emphasizes regions of the MPS essential for speech intelligibility. This procedure was tested using normal-rate and 2×-time-compressed sentences. The results indicated: (a) CImgs could be reliably estimated in individual listeners in relatively few trials, (b) CImgs tracked changes in spectrotemporal modulation energy induced by time compression, though not completely, indicating that "perceptual maps" deviated from physical stimulus energy, and