Abstract

Irino and Patterson (2002) have discussed the Mellin Transform as a candidate scale normalisation tool in the auditory system. What is the evidence for the usefulness of such a normalisation tool? In this report we review data on the formant frequencies present in human vowel sounds collected by Peterson and Barney (1952) to answer this question. We see how much vowel formant variability can be attributed to vocal tract length (VTL) changes. Interpreting the findings from the perspective of scale, we find a natural space in which to analyse scaled frequency components of sounds. By recasting Peterson and Barney’s data into this new representation we are able to quantify the usefulness of a scale normalisation tool. Finally we investigate the correlation between pitch and formant frequency, suggesting how these cues might relate to the percept of speaker size.

click here to open the movie

click here to open the report