Production and perception of an articulatory continuum for fricatives of English

Abstract Simple articulatory contrasts for a phonological opposition generate a multiplicity of acoustic cues. Knowledge of the covariations of acoustic pattern features within and across speakers is needed for automatic speech recognition. A model of speech production processes was used to generate stimuli along an articulatory continuum, the degree of abduction of the vocal folds for the fricatives in English words “hiss” and “his”. Transition times imposed strong constraints on the articulatory plans devised as inputs to the model. Acoustic segment durations output from the model covaried in the same way as those produced by 5 real speakers; there was good quantitative agreement in most cases. Listeners' responses suggest that the articulatory dimension synthesised is a suitable one for natural speech. Future directions for the modelling of multiple articulatory dimensions and for mapping from speaker-specific patterns of articulation and their perturbations onto the stability of particular acoustic cues are discussed.