Abstract: To realize joint source-channel decoding in the system with self-synchronization scrambling, a convolutional decoding like Soft-Input Soft-Output (SISO) descrambling algorithm is proposed in this paper by treating the self-synchronization scrambler as a special convolutional code. In this algorithm, residual source redundancy is used to update the extrinsic information of the scrambling sequence, which is then used as a prior information in the channel decoding process. So that, soft information exchanges with each other between the descrambler and the decoder, which efficiently improve the performance in the receiver. The performance of the proposed algorithm is verified by simulations. Results show that our approach outpeforms the conventional descrambler by 4.1 dB in the case of TPC encoder, while the residual source redundancy rate is 70%. Joint source-channel decoding in the system with self-synchronization scrambling can achieve larger gain compared with the system with only channel coding.