5th International Conference on Spoken Language Processing
The categorical perception (CP) of syllable-initial stop consonants has been intensively studied using psychophysical procedures over many decades. However, computational models consisting of an auditory `front end' and a learning system as a `back end' convincingly mimic the essentials of CP. Unlike real listeners, such models can be systematically manipulated to uncover the basis of their categorisations. In this paper, we explore the use of modern inductive learning techniques in simulating CP.
Bibliographic reference. Damper, Bob I. / Gunn, Steve R. (1998): "On the learnability of the voicing contrast for initial stops", In ICSLP-1998, paper 0843.