Joint source-channel coding and decoding is the hotspot in the field of coding recently. Realize joint source-channel coding (JSCC) with variable-length code, which can improve the transmission efficiency greatly.The paper adopts a kind of novel coding method, and constructs the cost-evaluating symmetrical reversible variable-length code that is fit for realizing joint source-channel coding and decoding. Finally the excellent simulation results have been get depending on the BSC and AWGN channels.Fix-length codes (FLCs) and variable-length codes (VLCs) both can realize JSCC, but the transmission efficiency of variable-length code is much higher than the other. One problem that how to solve the error diffusion taken by VLCs appears, and now we know Reversible variable-length codes (RVLCs) can resolve it exactly. But the traditional RVLCs has many problems on coding efficiency, error resilient capability and so on. So we propose a kind of novel RVLCs called Cost evaluating symmetrical reversible variable-length codes (C-E RVLCs). It is more sufficient than Takishima algorithm and makes the hanmming distance of the Takishima RVLC bigger. In this paper, we introduce the ultimatum, design rule and performance analysis of C-E RVLCs. Furthermore we propose the JSCC scheme and its simulation results.
C-E RVLCsSymmetrical reversible variable-length code (RVLCs) is the VLC with strong bit-error resilience, which can be decoded in both forward and backward directions. RVLCs received extensive attention especially during the development of the new video standards H.263+ and MPEG-4. Fraenkel and Klein proposed the necessary conditions of the RVLCs from the fixed code length. Takishima, Wada and Murakami proposed the first RVLCs codeword based on Huffman codes. Jiangtao Wen and Villasenor proposed novel asymmetrical RVLCs [2] that based on the exponential code length RLCs (Golomb-Rice code and exp-GR code).So RVLCs finally were adopted as the standards of MPEG-4.The reference [1] shows that the traditional RVLCs proposed by Takishima realized the transformation from Huffman codes to RVLCs, but the code-length vector is initialize by Huffman codes, which leads to the total average length of RVLC is bigger than Huffman code.