Technologies :

The coding group has chosen to focus its researches on the development of hybrid coders and, more particularly, on Code Excited Linear Prediction coders (originally proposed by Atal and Schroeder). These coders produce low-rate coded speech comparable to that of medium-rate waveform coders.

These coders determine the system parameters by linear prediction and the excitation sequence is determined by close-loop or analysis-by-synthesis optimization. The optimization process chooses an excitation sequence which minimizes a measure of the weighted difference between the input speech and the coded speech. The weighting function is chosen such that the coder is "optimized" for the human ear.

The synthesis block consists of a short-term synthesis filter which represents the formant structure of speech, a long-term synthesis filter which represents the pitch structure of the speech. A perceptual filter is used to shape the error such that the quantization noise is masked by the speech signal. A excitation generator forms or selects an excitation sequence such that the weighted Mean Squared Error (MSE) is minimized. The CELP coder produces low-rate coded speech comparable to that of medium-rate waveform coders.

Last updated December 17, 1999, send comments to dutoit@tcts.fpms.ac.be