 |
Workshop on the Auditory Basis of Speech PerceptionKeele University, UK |
 |
|
Workshop on the Auditory Basis of Speech PerceptionKeele University, UK |
|
Listeners were presented with pairs of concurrent vowels and requested to report one or two vowels. The dF0 was either 0 or 6%, and RMS levels before mixing were either the same or different by 10 or 20 dB. Responses for each vowel within a stimulus were classified according to relative level (-20, -10, 0, 10, 20 dB) and dF0(0 and 6%). Identification was better at dF0=6%, and this effect was greatest when the target was weak (-20 and -10 dB). This outcome is difficult to account for with current models, but can be explained by invoking a within-channel neural cancellation filter. A model of concurrent vowel identification based on this filter is consistent with our experimental data, and agrees with results that show that the auditory system segregates harmonic sounds by cancelling the harmonic background.
Bibliographic reference. Cheveigné, Alain de (1996): "A neural cancellation model of F0-guided sound separation", In ABSP-1996, 180-185.