11th Speech in Noise Workshop, 10-11 January 2019, Ghent, BE

Relating speech perception in noise to temporal-processing auditory capacities

Laurianne Cabrera(a)
CNRS – Université Paris Descartes, France

(a) Presenting

Temporal cues (e.g., amplitude modulation, AM) play a crucial role in speech intelligibility for adults. This study explored whether speech-in-noise (SIN) intelligibility relates more to sensory or cognitive factors involved in AM processing. AM masking and temporal integration were measured using non-speech sounds to better characterize sensory (AM encoding) and cognitive (memory and decision) mechanisms.

Twenty-two adults with normal-hearing completed three 3I-3AFC adaptive tasks. The first task assessed AM sensitivity using pure tone carriers and three modulation rates (4, 8, 32 Hz). The second task assessed AM masking by comparing AM detection thresholds at the same three modulation rates using three carriers varying in their inherent AM fluctuations: tones, narrowband noises with small inherent AM fluctuations and noises with larger fluctuations. The third task assessed temporal integration, the effect of increasing the number of AM cycles (between 2 and 8 cycles) on AM detection using tones modulated at 4 or 32 Hz. Finally, a fourth XAB task was designed to measure identification thresholds in speech-shaped noise using fricative and stop consonants. Four phonetic contrasts were tested here: changes in voicing or changes in place of articulation for either fricative or stop consonants.

Results showed that AM detection thresholds were affected by AM rate, large carrier fluctuations and number of AM cycles. Regarding SIN, thresholds were significantly better for changes in place of articulation for fricatives compared to stops and also compared to changes in voicing for fricatives. Preliminary regression analyses suggested that for stop consonants (either changes in voicing or place), better AM detection thresholds obtained in the different 3-AFC tasks contributed to explain up to 58% of the variance in the SIN thresholds. For fricative consonants, AM detection thresholds did not contribute to explain the variance in the SIN thresholds.

These results suggest that SIN perception is related to some extent to temporal processing, and better AM processing might contribute to better SIN perception for adults. Computational modelling will help to better disentangle the relationship between sensory and non-sensory processing of AM in SIN.

Last modified 2019-01-08 16:51:41