ARE THE PERCEPTUAL EFFECTS OF SPECTRAL SMEARING INFLUENCED BY SPEAKER GENDER? Academic Article uri icon

abstract

  • The purpose of this study was to determine whether the effect of loss of spectral detail on speech perception is influenced by the gender of the speaker. Spectral smearing was carried out by multiplying the speech signal by a series of low-passed white noise samples, causing tonal components in the signal to be replaced by noise. Smearing bandwidths of 0 Hz (no smearing), 250 Hz, 500 Hz, 1,000 Hz, 2,000 Hz, 4,000 Hz, 8,000 Hz and full bandwidth were used. Smearing was applied to 15 isophonemic lists, each with 10 one-syllable CVC Hebrew words. The words were recorded using two speakers, a male and a female, both native speakers of Hebrew. A total of 23 subjects participated in the study: eight listened to the male speaker and 15 to the female. The results show no significant differences in the effect of speaker on recognition of words, phonemes, vowels and consonants at the different smearing bandwidths. The results also show that regardless of the speaker's gender, vowels were adversely affected by spectral smearing, as compared to consonants. Interpolation of the results shows that smearing bandwidths of 1,080 Hz, 1,950 Hz, 1,590 Hz and 2,150 Hz are required to reduce word, phoneme, vowel and consonant recognition to 50%, respectively. Several tentative explanations are offered for the fact that the results were independent of gender: all smearing bandwidths were larger than the average interharmonic spacing for both speakers, and the difference between male and female formant frequencies is typically smaller then the difference in formant frequencies of the different vowels.

publication date

  • January 2004