Amplitude envelope cues for consonant recognition with simulated loudness recruitment

Abstract

The perceptual consequences of expanding the amplitude variations in speech to simulate loudness recruitment were studied under conditions in which spectral information was obscured to varying degrees, but audibility was largely preserved. Signal correlated noise samples, created individually to follow the amplitude envelopes of 22 VCV nonsense words, were used as maskers of those words, with S-N ratios of −15 to 0 dB. Amplitude expansion was by a factor of 3.0 in terms of decibels; e.g., a difference in levels of 10 dB between adjacent consonants and vowels was expanded to approximately 30 dB. Presentation level of speech peaks was 80 dB SPL. Many consonant levels were substantially less intense, but almost all were still audible. Recognition performance for the expanded speech by 50 normal-hearing listeners was as much as 30 percentage points poorer than unexpanded speech, and the types of errors were dramatically different, particularly in the midrange of S-N ratios. Follow-up testing as a function of level confirmed that differences between conditions were due largely to expanded amplitude envelopes rather than differences in audibility. The data suggest that amplitude envelope cues are vulnerable to the effects of expansion of the type seen in loudness recruitment. [Work supported by NIH ♯DC01533.]