On the Spectrum of Spoken English

Publisher Website
Google Scholar
Abstract
The spectrum of spoken English was measured in 13-octave bands for men, women, and children. The speech produced by reading aloud newspaper text was recorded in an anechoic chamber for each member of groups of ten men, ten women, and six children. The combined output of each of the three groups was obtained by electronically mixing the member outputs with a 16-track tape recorder. After mixing, the resulting recordings were analyzed in 13-octave bands and converted to spectrum levels. The spectra exhibit peaks and valleys not revealed with wider-band analysis, but when the results of the 13-octave analysis are combined to simulate octave bands there is agreement with the previously published literature. Next these free-field spectra were converted to give the spectra at the eardrum of a typical listener, thus allowing for head diffraction and the resonance of the ear canal. The spectra so obtained have peaks appropriate to typical values of Fo, F1, and F3 for each of the groups. The region usually associated with F2 spans a dip between the F1 and F3 regions. At the eardrum, the frequency region from about 1.7–4 kHz, which is usually associated with F3 and many of the friction generated components in speech, lies only 5–10 dB below the maximum levels in the low-frequency region.

This publication has 0 references indexed in Scilit: