Acoustic Correlates of the Voice Qualifiers: A Survey
This survey addresses the problem of understanding and utilizing voice quality for paralinguistic analysis, but it is incremental as it builds on existing research.
The paper maps paralinguistic voice qualifiers to their acoustic correlates and identifies openSMILE features for measurement, while providing examples of inferences like personality and health detection.
Our voices are as distinctive as our faces and fingerprints. There is a spectrum of non-disjoint traits that make our voices unique and identifiable, such as the fundamental frequency, the intensity, and most interestingly the quality of the speech. Voice quality refers to the characteristic features of an individual's voice. Previous research has from time-to-time proven the ubiquity of voice quality in making different paralinguistic inferences. These inferences range from identifying personality traits, to health conditions and beyond. In this manuscript, we first map the paralinguistic voice qualifiers to their acoustic correlates in the light of the previous research and literature. We also determine the openSMILE correlates one could possibly use to measure those correlates. In the second part, we give a set of example paralinguistic inferences that can be made using different acoustic and perceptual voice quality features.