Sixth European Conference on Speech Communication and Technology

Budapest, Hungary
September 5-9, 1999

Very Low Bit Rate Voice Coder Based on a Nonlinear Hearing Model

Rudolf Földvári, László Gyimesi

Széchenyi István College Györ, Hungary

A new hearing model based on the knowledge of anatomy and phychoacoustical phenomena was presented by R. Földvári and Gy. ács in 1996. After a short survey of older hearing models, they proved that their hearing model is able to describe true hearing experience (critical bandwidth, phase limit frequency). We suppose that the outputs of Zwicker's filters and the nonlinear transformation secured by the nonlinear hearing model is suitable for brain processing. Furthermore, we wish to point out that by realizing this method a very low bit rate vocoder and in addition an efficient speech enhancement can be achieved in cases when signal-to-noise ratio is about 0 dB.

Full Paper (PDF)   Gnu-Zipped Postscript

Bibliographic reference.  Földvári, Rudolf / Gyimesi, László (1999): "Very low bit rate voice coder based on a nonlinear hearing model", In EUROSPEECH'99, 1547-1550.