Formant-based articulatory normalization and its application to vowel restoration

Abstract

Visual feedback of spontaneous speech is effective for articulatory training in deaf children and speech rehabilitation in patients with dysarthria. Particularly, visual representations of vowel formant frequencies have been used directly and indirectly for such purposes because these acoustical parameters reflect the articulatory behavior. However, since not only the shape but also the size of the vocal tract affects the formant frequencies, minimization of the effects due to differences in size of the vocal tract is required. In such speaker normalization, we define a color space consisting of three ratios of formant frequencies and apply it to the color visualization of vowels. In this paper, we propose normalized articulation space as an expansion of color space, where we assume that the neutral vowel of any speaker is mapped onto a unique point. In addition, since the proposed articulatory space is regarded as a speaker-independent representation of the vocal tract shape, we also propose a method for converting the modified articulation plane into formant space for the purpose of correcting degraded speech.