Estimation of speaker's height and vocal tract length from speech signal

Estimation of speaker’s height and vocal tract length (VTL) from speech signal can have forensic and automatic speech recognition applications. It was suggested for a long time that there is a correlation between speaker’s VTL, on one side, and speaker’s height and formant frequencies, on another side. Until recently, these putative relationships have been empirically examined in studies employing relatively small numbers of speakers. Scattered studies presented intriguing results about the correlations between speaker’s height and various acoustic speech parameters. Due to lack of databases, few studies presented extensive comparative results between the actual speaker’s VTL and the estimated one from speech signal. This paper presents an analysis of correlations between various acoustic speech parameters and speaker’s height for a large number of speakers. It also presents a new method for an optimal estimation of speaker’s height and VTL from various acoustic speech parameters.