Quast, Holger
Coauthors(s):
UCSD
Machine Perception Lab, Institute for Neural Computation
INC, UCSD 9500 Gilman Dr. Dept. 0523 La Jolla, CA 92093-0523

Absolute Perceived Loudness of Speech

Humans have the ability to recognize a speaker as loud or quiet disregarding how strong the actual stimulus is at our ears. We can tell if a person is screaming or speaks barely louder than whispering, whether he/she is amplified by a rock concert size 40kW PA system or whether we hear the speaker through a barely audible radio in the background. Previous studies have measured vocal effort by means of inverse filtering and glottal pulse slope analysis. However, even if a person exercises high vocal effort, this does not necessarily imply he/she is perceived as loud. I will present a psychoacoustically motivated model for perceived speech loudness that is invariant to amplifier gain settings. One use for this model is in nonverbal speech recognition and modeling the perception of para- and extralinguistic speech, where so far mostly quantities like power or energy have been used to describe a loudness dimension. Using only power neither permits to make assumptions about absolute loudness, nor, even worse, is it a valid representation of relative heard loudness, since the characteristics of our hearing are ignored. This suggests that the model to be presented, since it allows to evaluate not only relative perceived magnitudes like power or energy in one recording, but also the comparison of loudness values among speech samples with different speakers and recording settings, is a valid extension to this field.