Abstract: Since Variable-Length Codes（VLCs） are inherently vulnerable to channel errors，traditional error-correcting codes are unable to solve the propagation-error problem efficiently.A new symbol-constrained joint source-channel decoding algorithm for VLCs is presented.By constructing double-tree VLCs sources，the propagation-error problem is evidently alleviated.Meanwhile，by applying symbol-constrained method to MAP algorithm，not only the a priori probability of VLCs sources is efficiently utilized，but also the Symbol Error Rate（SER） is reduced enormously.Compared with traditionally bit-constrained MAP algorithm，simulation results demonstrate that for JPEG source over AWGN channels，the gain in SNR of the method is around 0.8dB when SER reaches 1%.

摘要： 由于不等长编码信源（VLCs）所固有的易于误码扩散的弱点，传统的纠错编码并不能高效地解决其差错控制问题。提出了一种新的针对不等长编码信源的符号约束MAP联合译码算法，利用VLCs双树信源的构造方法，有效地抑制了VLCs的错误扩散；同时，通过将“符号约束”的思想应用于最大后验概率（MAP）译码算法，不仅充分利用了信源先验信息，而且极大地降低了误符号率。实验结果表明，在误符号率为1%时，对于在高斯加性白噪声（AWGN）信道传输情况下的JPEG信源，提出的算法比传统的“比特约束”MAP译码算法有将近0.8 dB的性能增益。