Critical speech bands: Spectral resolution of contrasting temporal speech patterns

Abstract

The high intelligibility of narrow-band sentences is attributable to contrasting temporal patterns in different frequency regions within the narrow band [E. W. Healy, Dissertation, University of Wisconsin–Milwaukee (1998)]. In the current study, the spectral resolution for contrasting temporal speech patterns was measured by partitioning the narrow-band spectrum. Sentences from the HINT were filtered to a minimum overall bandwidth, centered at 1500 Hz, which allowed relatively high intelligibility of component words. The overall bandwidth was 2/3, 1, or 3/2 octaves. Each overall band was digitally filtered (several thousand dB/octave) into 1, 2, 4, 6, or 10 partitions. The envelope from each partition was used to amplitude-modulate a corresponding band of low-noise noise, and the modulated carriers were combined and presented to normal-hearing listeners. Intelligibility increased and reached an asymptote as the number of partitions increased. The number of partitions necessary for asymptotic performance was 4, 6, and 10 for the 2/3-, 1-, and 3/2-octave overall bandwidths. For each, this corresponds to a partition bandwidth of approximately 1/6 octave. These results indicate that, in the region surrounding 1500 Hz, resolution of individual temporal speech patterns approximates the critical band measured using traditional psychoacoustic methods and nonspeech stimuli. [Work supported by NIDCD.]