Compressed Domain Noise Reduction and Echo Suppression
D.J. Marchok · 2000
In this paper, a novel approach to speech processing in the compressed &main is discussed. The compressed domain refers to the quantized representation of the speech signal in terms of a set of parameters using a particular compression method. The coded speech may require only partial decoding to extract the information necessary to facilitate enhancement Even with full decoding, the additional processing cost is low since decoding is usually an order of magnitude less expensive than encoding. The key to the savings in processing cost, when using the compressed domain speech enhancement techniques, lies in the partial re-encoding of only a few of the speech parameters. The cost of this partial encoding can be several oders of magnitude lower than full encoding. Compressed domain techniques for noise reduction and echo cancellation are discussed in this paper.