Gaps in the Acoustic Theory of Speech Production

Abstract

Speech technology has demonstrated impressive applications such as low bit-rate speech coding, text-to-speech-synthesis, automatic speech recognition and speaker recognition etc. Such technological applications rely on speech analysis which in turn is based on a model for speech production. A success in these applications gives rise to an impression that we have a scientifically sound model for speech production. In this paper we review the articulatory-acoustics and the source-filter models of speech production and show that there exist gaps in our present knowledge. We propose a holistic approach of speech signal analysis for future research.