A Voice Production Model for Waveform Coding Speech

Voice production models are used for speech compression thus reducing the encoding data rates and storage space for speech signaling processing. The legacy voice production model consists of a vocal source model and a vocal tract model. This paper presents a simpler voice production model adapted to waveform coding of speech: the Mathieu Waveform Coder (MWC) model. The MWC model yields weighted basis functions to waveform code the speech signal. Each Mathieu basis function represents an infinite series of sinusoids rather than a single sinusoidal spectral line. The research motivation is to determine if waveform coding speech using an elliptical membrane model (i.e., the Mathieu basis functions) is more efficient than coding speech using a circular membrane model (i.e., the sine basis functions) . Preliminary results indicate that waveform coding using the MWC model provides a superior estimate of the speech signal compared to coding with an equal number of sinusoids.