5aPP11. Estimation of nonsense syllable intelligibility.

Session: Friday Morning, December 6

Time: 10:45

Author: Christine M. Rankovic
Location: Dept. of Speech Lang. Pathol./Audiol., Northeastern Univ., Boston, MA 02115


An intelligibility estimation test comprising nonsense syllables that were recorded separately, and then concatenated, is under development for hearing aid assessment. Consonant-vowel-consonant syllables (CVCs) are presented along with a visual display on a computer monitor that identifies each syllable with an animated pointer as it is presented to the listener. After presentation of a list of 50--60 CVCs, the listener is asked to estimate the consonant percent-correct score. A major advantage of the test is rapid administration. Percent-correct estimates were collected from 16 normal-hearing subjects presented with CVCs masked by a speech-shaped noise and with high- and low-pass filtered CVCs. In all test conditions, actual percent-correct scores were obtained for comparison. In noise and under filtering, estimates are slightly higher than actual scores. However, estimates increase with speech-to-noise ratio (dB) at a rate of about 4% per dB, similar to the actual percent-correct scores. The crossover frequency derived from the filtering experiment was approximately 1700 Hz, nearly identical to that reported by French and Steinberg [J. Acoust. Soc. Am. 19, 90--119 (1947)]. Hence, articulation theory principles may be applicable. This test may serve as a speedy assessment tool for evaluating and selecting hearing aids. [Work supported by NIH.]

ASA 132nd meeting - Hawaii, December 1996