| Summary: | 碩士 === 國立高雄第一科技大學 === 電腦與通訊工程所 === 97 === Concatenated codes are a class of powerful error-correction codes. It has been widely used in a variety of communication systems such as mobile communications, DVB-T2 broadcasting standard, and the CCSDS space communication standard. A typical concatenated code consists of a random-error correcting convolutional code and a burst-error correcting block code which are decoding with a Viterbi algorithm and an algebraic decoder, respectively. The error performance requirement is usually satisfied, however, the coding gain is still not fully exploited due to lack of maximum likelihood (ML) decoding. If ML decoding is possible, many existing communication systems can improve their performance further. Since the invention of turbo codes, high-performance turbo signal processing becomes a paradigm of new communication systems. How to improve the performance of concatenated codes via turbo decoding is an interesting research topic.
Traditional turbo codes consist of two identical component codes such as turbo convolutional codes and turbo block codes. The researches about hybrid-component turbo codes are rare, that is the open problem of ML decoding for concatenated codes. In this thesis, we will focus on the study of hybrid turbo codes (HTC). Many important issues of serial HTC (SHTC) and parallel HTC (PHTC) are studied in detail including the encoding/decoding architectures, decoding methods, performance and analysis. In SHTC, we propose two new decoding methods and one of the proposed methods outperforms the traditional decoding method. Both the BCJR and SOVA are employed. As expected, the BCJR is better than SOVA for all kind of decoding methods. We extend the extrinsic information transfer chart (EXIT Chart) to the analysis of HTCs and the performances prediction of EXIT chart is consistent with the simulation results. For AWGN channels and BER = 10-5, the capacity gap of the proposed PHTC and SHTC are 1.5dB and 2dB away from Shannon’s limits, respectively.
|
|---|
