Abstract: In this dissertation, we focus on three issues of the trellis based iterative decoding: First, the complexity issue of Turbo code is considered. We propose a constrained iterative decoder to reduce the decoding complexity. An additional interleaver is introduced at the encoder. At the decoder, we first use Cyclic Redundance Code (CRC) to detect which bits are already correctly decoded during early iterations. With knowledge of the positions of these correct bits, the constrained decoding algorithm is designed to reduce the number of the state transitions in the component code trellis and help the decoding of other bits in later iterations. In this way, the constrained iterative decoder achieves significant complexity reduction and still satisfying performance. Second, the iterative decoding algorithm is redesigned for Turbo code implemented Distributed Source Coding (DSC). When used in DSC, the Turbo decoder encounters a combined Binary Symmetric Channel (BSC) and Addictive White Gaussian Noise (AWGN) distortion. The existing iterative decoding algorithm based on AWGN distortion assumption causes performance degradation. By redefining the channel reliability values, the modified iterative decoding algorithm matches the BSC-AWGN scenario well and improves the performance. Third, we propose a reliable source transmission coding and decoding scheme. A serially concatenated source and space time modulated coding structure is used. Variable Length Code (VLC) with error resilient capability is adopted at the application layer. Space Time Trellis Code (STTC) is used to provide high bandwidth efficiency at the physical layer. An iterative joint source space time decoder is designed including the symbol level space time Maximum A Posteriori (MAP) decoder, the bit level VLC MAP decoder and the Viterbi VLC decoder. Critical issues such as STTC MAP algorithm with nonseparable systematic information, VLC MAP algorithm in absence of channel output, VLC Viterbi algorithm based on the bit level trellis and extrinsic information conversion and exchange between bit domain and symbol domain are addressed. The decoding performance of different frame sizes and different component VLCs and STTCs, the rate allocation between the source code and the space time code and the performance in presence of channel estimation errors are discussed in this dissertation.