B. Espinoza-Varas
Communication Sciences & Disorders, Univ. Oklahoma Health Sciences Ctr., Oklahoma City, OK 73190
Muralidhar R. Kudumala
Univ. of Oklahoma, Oklahoma City, OK 73190
The masking of additive noise by vowel sounds is relevant to speech coding. Since the spectral density of vowels is relatively low, and varies inversely with fundamental frequency, it may limit the masking effectiveness of vowels. In this paper, an excitation-pattern (EP) approach (Moore and Glasberg, 1987) is used to model effects of spectral density on masking. EP predictions were obtained for masker densities ranging from 1.0 spectral component per Hz to 1.0 per 220 Hz, and target densities ranging from 1.0 component per Hz to 1.0 per 220 Hz. In all conditions, the overall power of masker and targets was 64.0 and 50.3 dB, respectively. For each combination of masker and target density, the EPs of the masker alone (EP[inf m]) and of the masker-plus-target (EP[inf m+t]) were obtained. The EP difference, EPD=(EP[inf m+t])-(EP[inf m]), at each ERB was integrated (across all ERBs) to obtain an index of target detectability. For masker densities lower than ~1.0 component per 100 Hz, detectability improved as masker density decreased; the effect of target density was small. Predictions are compared to detection thresholds for additive noise masked by either /i/ or /a/, with 100- or 200-Hz fundamentals. [Work supported by OCAST-HR4-064.]