Voice Quality: Functions, Analysis and Synthesis August 27-29, 2003 Geneva, Switzerland

On the Representation of Voice Source Aperiodicities in the MBE Speech Coding Model

Preeti Rao, Pushkar Patwardhan

Department of Electrical Engineering, Indian Institute of Technology, Bombay, India

We present an investigation of the representation of voice source aperiodicities in the Multi-Band Excitation (MBE) speech model for the compression of narrowband speech. The MBE model is a fixed-frame based analysis-synthesis algorithm which combines harmonic and stochastic components to reconstruct speech from estimated model parameters. Pitch cycle perturbations, such as jitter and shimmer, are not captured accurately in the framewise constant parameter estimates, thus impacting the reproduced voice quality. The actual dependence of MBE reconstructed voice quality on the voice pitch and the type of perturbation are explored through objective measurements and subjective listening with synthetic and natural speech.

Full Paper    Presentation (PDF; 1757 KB)    Presentation (Powerpoint; 3038 KB)

Bibliographic reference.  Rao, Preeti / Patwardhan, Pushkar (2003): "On the representation of voice source aperiodicities in the MBE speech coding model", In VOQUAL'03, 121-126.