Very loud speech over simulated environmental noise tends to have a spectral peak in the F1 region

Abstract

In some professions, workplace noise appears to be a hazard to the voice, if not to hearing. Several studies have shown that teachers and sports instructors, for example, are more prone to voice problems than average, prompting research on loud voice. Since on-location recordings are in many ways impractical, the running speech of 23 untrained speaker subjects (12 female, 11 male) was instead recorded in several types of loud noise that was presented over high-quality loudspeakers. Using adaptive cancellation techniques, the noise was then removed from the recordings, thus exposing the strained voices for analysis. The experiment produced a large body of data, only one aspect of which is reported here. In most subjects, the vowel spectrum as a function of voice SPL showed the expected behavior for low to moderate efforts, but developed a very pronounced peak in the F1 region at the highest efforts. This peak can be ascribed to the concerted action of several acoustic mechanisms, including source waveform asymmetry, F1 approximating one of the lower partials, and increased formant Q values due to a longer closed phase. [Work supported by the Swedish Council for Working Life and Social Research, Contract No. 2001-0341.]

Effect on LTAS of vocal loudness variation

“…Under such conditions, a different relationship between Leq and LTAS levels can be expected. Therefore, we assume that the linear relationship between Leq and LTAS level apply to a normal range loudness variation range (Ternström, 2003).…”