Magnetic Resonance Imaging of the Vocal Tract - Techniques and Applications

Abstract

Magnetic resonance (MR) imaging has been used to analyse and evaluate the vocal tract shape through different techniques and with promising results in several fields. Our purpose is to demonstrate the relevance of MR and image processing for the vocal tract study. The extraction of contours of the air cavities allowed the set-up of a number of 3D reconstruction image stacks by means of the combination of orthogonally oriented sets of slices for each articulatory gesture, as a new approach to solve the expected spatial under sampling of the imaging process. In result these models give improved information for the visualization of morphologic and anatomical aspects and are useful for partial measurements of the vocal tract shape in different situations. Potential use can be found in Medical and therapeutic applications as well as in acoustic articulatory speech modelling. Magnetic Resonance (MR) improvements, in the past decades, allowed vocal tract imaging, making it currently one of the most promising tools in speech research. Speech is the most important instrument of human communication and interaction. Nevertheless, the knowledge about its production is far from being complete or even sufficient to describe the most relevant acoustic phenomena that are conditioned at morphological and dynamic levels. The anatomic and physiologic aspects of the vocal tract are claimed to be essential for a better understanding of this process. The quality and resolution of soft-tissues and the use of non-ionizing radiation are some of the most important advantages of MR imaging (Avila-García et al., 2004; Engwall, 2003). Several approaches have been used up to now for the study of the vocal tract based on MR images. Since the first study proposed by Baer et al. (1991), many MR techniques have been used (from static to dynamic studies, and more recently even done in real-time), starting by studies of vowel production (Badin et al., 1998; Demolin et al., 2000), followed by consonant production (Engwall, 2000b; Narayanan et al., 2004), for different languages such as French (Demolin et al., 1996; Serrurier & Badin, 2006), German (Behrends et al., 2001; Mády et al., 2001), and Japonese (Kitamura et al., 2005; Takemoto et al., 2003). The work presented in this paper, consisting basically in the static description of the vocal tract shape during sustained vowels and consonants and in the dynamic description of some syllables, is the first to report the application of MR imaging for the characterization of European Portuguese (EP). This study started in 2004, having attained a first series of results published in 2006 (Rua & Freitas, 2006). Our approach can be seen as a contribution to the wide area of articultory speech modeling, since it provides geometrical data to the acoustic modeling phase or research. In the articulatory speech research of EP a few studies of nasal vowels have been carried through, at the acoustic production and perceptual levels based on acoustic analysis and electromagnetic articulography (Teixeira et al., 2001, 2002, 2003). More recently, another MR study of EP presents some results relative to oral and nasal vowels exploring contours extraction from 2D images, articulatory measures and area functions (Martins et al., 2008). In former studies, vocal tract modelling has been limited to the midsagittal plane (Engwall, 2000a; Takemoto et al., 2003), but improvement of MR imaging equipment system capabilities allowed the expansion into this domain of research and made it possible to obtain three-dimensional (3D) modelling (Badin & Serrurier, 2006). The more realistic models of the vocal tract shape that nowadays are possible to obtain, are hugely needed in the research towards improved speech synthesis algorithms and more efficient speech rehabilitation. The main purpose of this paper is to present some 3D models of the vocal tract based on MR data of some relevant sustained articulations of EP in a static study. From the point of view of image processing, a new approach for 3D modelling by means of the combination of orthogonal stacks, to describe the vocal tract shape in different articulatory positions is presented. We also demonstrate an MR technique to capture useful image sequences during speech (dynamic study). In addition, some preliminary results of this dynamic study are presented. The remaining of this paper is organized in four sections. The next section is dedicated to the methods and describes the equipment, corpus and subjects, as well as the procedures used for the speech study, namely for morphologic and dynamic imaging of the vocal tract. The results are presented in following section, through the exhibition of some three-dimensional models built of the vocal tract and an image sequence obtained during speech. Finally the conclusions of the work described are presented.