Speech Coding in Noisy Environments based on Human Auditory System Model

In this paper، an analysis/synthesis system based on the natural Cochlear model and the human auditory perception properties is presented that is capable of coding the speech signal under harsh acoustic conditions. For this purpose، the corrupted signal is decomposed into a number of sub-bands using complex Gammatone filter bank، and the sub-band signals are individually and adaptively denoised. Temporal masking، a combination of a novel non-uniform quantization method and lossless coding algorithms are then applied to reduce the number of coefficients and compression. The proposed system is very robust and effective in dealing with different kinds of noises such as white، colored and periodic noises. Objective and Subjective quality measurements demonstrate significant improvement in the perceptual quality of synthesized speech while considerable reduction in bit-rate to approximately 14. 6 Kbps is achieved.