SOURCE-CONTROLLED CHANNEL DECODING

Source and channel coding have been treated separately in most cases. It can be observed that most source coding algorithms for voice, audio and images still have correlation in certain bits, Transmisson errors in these bits usually account for the significant errors in the recon structed source signal. This paper proposes a modification of the Vite rbi decoding algorithm (VA) for binary trellises which uses a priori o r a posteriori information about the source bit probability for better decoding in addition to soft inputs and channel state information. An alytical upper bounds for the BER of convolutional codes for this modi fied VA (APRI-VA) are given, The algorithm is combined with the Soft O utput Viterbi algorithm (SOVA) and an estimator for the residual corre lation of the source bits to achieve source-controlled channel decodin g for framed source bits. The description is simplified by an algebra for the log-likelihood ratio L(u)= log(P(u = +1)/P(u = -1)) which allo ws a clear definition of the ''soft'' values of source-, channel-, and decoded bits as web as a simplified description of the traceback vers ion of the SOVA. Applications are given for PCM transmission and the f ull rate GSM speech codec. For an PCM coded oversampled bandlimited Ga ussian source transmitted over Gaussian and Rayleigh channels with con volutional codes the decoding errors are reduced by a factor of 4 to 5 when the APRI-SOVA is used instead of the VA, This results in much le ss signal distortion. The GSM decoder is modified for those few signif icant bits only which still have correlation between consecutive 20 ms speech frames, A simple dynamic Markov correlation estimator is used, With these receiver-only modifications the channel SNR in a bad mobil e environment can be lowered by 2 to 3 dB resulting in the same voice quality, Further applications are briefly discussed.