Spectral interpolation coder for impulse noise cancellation over a binary symetric channel

In this paper, a spectral interpolation coder (SIC) and decoder are investigated for simultaneous source coding and impulse noise cancellation. For simplicity of the analysis, we restrict ourselves to the framework of scalar quantization of a memoryless gaussian source to be transmitted over binary symmetric channel (BSC). Our approach is to make a carefully designed interpolation of the data in the spectral domain prior to quantization and transmission. The SIC decoder then exploits the properties of SIC codes in order to analyse, detect and correct (or reduce) erroneous data. A nice feature of this procedure is that the decoder deals simultaneously with the quantization noise and impulse channel noise; therefore it is able to reduce distortion introduced not only by the transmission channels errors but also by the quantizer. A comparaison study is also investigated in this paper: Simulations show that we obtain a 3dB improvement in SNR over the classical TSC scheme for a global rate of 8.2 transmitted bits per source sample and small BSC crossover probability.