Development and testing of artificial low-frequency speech codes
Charlotte M. Reed, Matthew H. Power, Kristin K. Foss, N. I. Durlach, L. D. Braids · The Journal of the Acoustical Society of America · 1987
Previous work on frequency lowered speech for listeners whose hearing is confined to low frequencies has focused on signal processing of natural speech. In the present research, the use of artificial low-frequency codes for such listeners is explored. This research will help separate limitations imposed by signal processing schemes from limitations imposed by the listener's perceptual system. It will also provide background for the design of new systems based on the use of speech-recognition devices as front ends. Artificial frequency-lowering codes were developed for 24 consonants (C) and 15 vowels (V) for two values of low-pass cutoff frequency F (300 and 500 Hz). Individual phonemes were coded by a unique acoustic signal confined to frequencies less than or equal to F. The ability of normal-hearing listeners to identify coded C's and V's in fixed-context syllables was compared to their performance on single-token sets of natural speech utterances low-pass filtered to equivalent values of F. Performance on coded C's was superior to that on low-pass filtered speech for both values of F (by roughly 20 percentage points). An advantage for coded V's (of roughly 15 percentage points) was observed only at F = 300 Hz. Identification of C and V components in roving-context CV syllables was generally inferior to that obtained in the fixed-context conditions. [Work supported by NIH.]