Relating LPC modeling to a factor-based articulatory model
Sacha Krstulović · 2000
This paper proposes a method for recovering the articulatory parameters of a factor-based vocal tract shape model from the speech waveform. This is realized by analytically relating the shape model to a Linear Prediction lattice lter. Results pertaining to human vowels are presented. They show a good agreement with phonetic characteristics in a real-time computational framework. 1. INTRODUCTION Most of the existing articulatory models use an area function (describing a discrete tube shape) as the interface between the articulatory level and the acoustics of speech. Alternately, it is well known [MG76, Wak79] that the process of Linear Prediction (LPC) is equivalent, under certain hypotheses, to acoustic ltering in discrete tubes. This equivalence has already been exploited to recover area functions from acoustics alone [Wak79]. In the present article, this method is extended by two further steps, namely: projecting the area function into the space of sagittal cuts through the ana...