Acoustic analysis of initial consonants in the California Syllable Test

Abstract

The goal of the present study is to conduct an acoustic analysis of onset consonants to identify the spectrotemporal features that distinguish them from each other and to identify acoustic consonant variations that produce the observed patterns of perceptual confusions seen in young- and older-normal-hearing listeners [J. Acoust. Soc. Am. 127, 1609–1623 (2010); JRRD 49, 1277–1292 (2012)]. We used the California Syllable Test (CaST) token set, which includes 40 exemplars of each initial consonant for each of three vowels and six talkers. The CaST measures recognition of 20 initial and 20 final consonants in speech-spectrum noise with each consonant presented at a 67%-correct signal-to-noise ratio (SNR). Time-normalized spectrograms are computed for each exemplar (from consonant onset to the end of the formant transition) by varying the time-spacing of the FFT spectral lines in proportion to the exemplar duration. Quantitative comparisons among normalized spectrograms of correctly recognized and confused exemplars at a range of SNRs suggest explanations for specific consonant confusions. The long-term goal is to apply this analysis to understand the effects of hearing loss and HAs on consonant perception and to predict consonant confusion patterns obtained with the CaST in normal-hearing and hearing-impaired listeners.