Compressed domain speech enhancement based on the joint modification of codebook gains

A compressed domain speech enhancement method based on the joint modification of adaptive and algebraic codebook gains for the codec of ITU-T G.722.2 is proposed in this paper. First the power of excitation signal corresponding to the noise is estimated by the method of minimum statistics. Then the decision-directed approach is used to get an estimate of the a priori SNR. And the algebraic codebook gain is modified by multiplying a Wiener-type modification factor. In order to solve the problem of power loss in voiced segment, the modified adaptive codebook gain is got by keeping the power of the modified excitation signal equal to the scaled version of the noisy one. The result of performance evaluation under ITU-T G.160 shows that, in comparison with the method that only modifies the algebraic codebook gain, the proposed method could provide larger amount of noise reduction in both white and colored noise with smaller attenuation on the speech level, and the objective speech quality is improved evidently.