Vocal effort levels in anechoic conditions

摘要

This paper presents statistical data on vocal effort levels in anechoic conditions from 50 British English speakers. Extending and rectifying the broadly quoted work carried out by Pearsons et al. in 1977 with American English speakers, the current work stipulates five vocal effort labels: hushed, normal, raised, loud and shout using a set of more precise descriptors with an aim to (1) avoid ambiguity and mis-interpretation of these effort levels, (2) extend the Pearsons' dataset to include a hushed speech level, which is demanded for the investigation of speech intelligibility and privacy, and (3) validate Pearsons' dataset with native British English speakers. In addition, the statistical data also show a high level of correlation between peak levels and equivalent continuous levels, providing a useful means to estimate speech levels in the presence of non-trivial background noise.