Phenomenological Model for Vowel Production

Comments on:
"A Phenomenological Model for Vowel Production in the Vocal Tract," Herbert M. Teager, and Shushan M. Teager
==================================================================

The major focus of alternative or a definitive model for speech (in particular, for Vowel production) is to remove the shackles which bind the uncodified and unexplored yet seemingly solved problem of speech because the technology has bypassed the need to do so.

Some scientests believe that 90% of the observations for speech can be explained with source filter theory, but the remaining 10% is not appropriately addressed.

The speech models tacitly depend on the validity of models for hearing. ****

The current model for speech - linear filter model - views voice produced as combination of pure-tones, with ear (imperfect Fourier analyzer) is believed to extract the magnitude of these objective components.

Linear model cannot predict or test / verify the transients / noise tones existing in nature.

The observations made by Teager (to highlight the anamolies in the modeling world) are:

A) the blame of limited and constrained speech systems (and its performance) is laid on our limited understanding of human brain. Teager examplifies by stating that a bird (myna) can replicate (mimic) the long passages of human speech inspite of having anatomically small cochlea (cochlea is assumed to be a place where in humans differentiate the tones into different frequencies, the different hair cells in the ear respond different to the set of frequencies, with certain set (of hair cells) responding to a particular set of frequencies), a different vocal tract (as compared to humans) and a different cerebral cortex structure to mitigate the speech signal between ear and spoken speech.

B) The various clicks, whistles, snores and other types of sounds that humans can produce cannot be modelled by linear conventional source filter model (as they do not assume the source to be glottis).

C) In theory, formant values depends solely upon the cross-sectional areas along the center line of the supraglottal vocal tract