A Novel Method for Wavelet Quantization of Noisy Speech

This paper proposes an architecture for low bit rate coding of noisy speech. The input noisy speech is decomposed into multiresolution signal components using wavelet transform. An iterative Wiener filtering is used at each level of wavelet analysis to enhance speech. The system model that evolves during enhancement is processed further to get optimal parameters for the quantization. A multistage vector quantizer is used for compression of decomposed speech. The enhanced speech is reconstructed at the receiving end by a VQ decoder and the necessary wavelet reconstruction network. Speech coding rate for the proposed architecture is estimated to be about 2.37Kbps.