Dependence of long‐term‐average spectrum curves of voices as function of equivalent sound level.

Abstract

Long‐term‐average spectrum (LTAS) is a quick and simple analysis method, which is frequently used in voice analysis. It typically reaches a stable curve after about 30 or 40 s of speech or singing and reflects both voice source and formant frequency characteristics of a voice. However, it is quite sensitive to changes of vocal loudness. In two previous investigations a set of recordings were analyzed in which 15 male and 16 female adults read a standard text aloud at seven different degrees of vocal loudness. The results showed that the LTAS curve changed linearly but in a frequency dependent manner with the equivalent sound level Leq. Furthermore, the alpha ratio between the summed energy below and above 1 kHz was found to be a quadratic function of Leq. In the present investigation we study the relation between Leq and (1) the overall LTAS slope between 0.75 and 4 kHz and (2) the average increase in LTAS level, in dB, in the same range. While slope varied considerably between speakers, average LTAS level increase could be accurately described by a linear function of Leq which is similar between speakers. Possible applications of these findings will be discussed.