Speech perception: Ideas and concepts initiated by Reinier Plomp

Abstract

In the late 1960s, a central issue in the research group on speech and hearing at TNO, headed by Reinier Plomp, was the relation between the perceptual and physical characteristics of sound. One impressive example concerned the research on vowel sounds, conducted together with Louis Pols. By adapting novel approaches for studying supra-threshold phenomena (triadic comparison of stimuli, Kruskal’s multi-dimensional scaling technique), the perceptual differences among vowel sounds could be visualized in a two- or three-dimensional representation of vowel points. This representation appeared to be in excellent agreement with a physical configuration of the vowel points, based on one-third-octave band analysis of the vowel sounds. It is conceptually only a small step from these isolated steady-state vowels, corresponding to points in a multi-dimensional space, to running speech resulting in a trace within that space. The spectral distances between successive points in such a speech trace reflect the rate-of-change in the speech signal as perceived by a listener. This framework of thinking, i.e., the perceptual relevance of projecting speech in the spectral domain, did lay at the roots of our later work on the development of objective measures for quantifying speech intelligibility.