Long-term average speech spectra for normal and hearing-imparied adolescents

Acoustical aspects of the speech of hearing-imparied persons have usually been examined over brief periods of time for simple utterances. Of interest in this study was whether abnormal acoustical patterns would be observed when speech was examined over longer periods of time. Spectra were averaged for 20 hearing-impaired adolescents, for whom we had audiometric profiles and an assessment of speaker intelligibility, and also for ten normal-hearing adolescents. An analyzer having constant-frequency bandwidths of 60 Hz, with the center frequencies of the filters spaced at 40-Hz intervals across a 10-kHz range, was used in the measurements (averaging time 27 s). The normal spectra are characterized by a regular pattern of peaks occurring at multiples of the talkers' fundamental frequencies and by slopes declining at rates of -5 to -6 dB/octave. After correction for lip-radiation impedance, these slopes are similar to that reported for the normal glottal source ( - 12 dB/octave). The hearing-impaired adolescents produced spectra for which the harmonic structure ranged from the very well defined to the irregular and poorly defined; spectral slopes declined at rates equal to or greater than the normal rate, in some cases declining at twice the normal rate.