In many realistic communication environments, experiments supported by theoretical considerations indicate the overall noise statistics deviate from the Gaussian distribution due to the presence of high amplitude interference. The decoding in non-Gaussian noise is complicated by the fact that accurate noise statistics are typically unavailable at the receiver. Without exploiting the noise probability density function (pdf), the widely accepted method is to erase the symbols corrupted by the high amplitude interference before the error-and-erasure decoding is performed. The accuracy of the erasure marking determines the decoding performance. In this thesis, we propose a joint erasure marking and decoding (JED) approach, which exploits the code structure in erasure marking such that the e...[ Read more ]

In many realistic communication environments, experiments supported by theoretical considerations indicate the overall noise statistics deviate from the Gaussian distribution due to the presence of high amplitude interference. The decoding in non-Gaussian noise is complicated by the fact that accurate noise statistics are typically unavailable at the receiver. Without exploiting the noise probability density function (pdf), the widely accepted method is to erase the symbols corrupted by the high amplitude interference before the error-and-erasure decoding is performed. The accuracy of the erasure marking determines the decoding performance. In this thesis, we propose a joint erasure marking and decoding (JED) approach, which exploits the code structure in erasure marking such that the erasures can be marked more accurately. The JED is investigated from the theoretical aspects to the application aspects.

Firstly, the theoretical aspects are studied. It is proved that the decision re-gion of JED is non-convex and decision boundaries are formed by hyperplanes and quadratic surfaces. The complex shape of the decision regions makes the per-formance evaluation difficult. Nevertheless, an error rate lower bound is obtained by considering the dominating terms.

We then implement the JED in trellises and derive the joint erasure marking and Viterbi algorithm (JEVA). We show that the JEVA is a maximum likelihood decoding scheme which finds the most likely transmitted code sequence with a set of symbol erasures without noise pdf. The performance of JEVA approaches that of the optimal maximum likelihood decoder that exploits the exact noise pdf. To trade off the computation complexity, decoding delay and memory re-quirement, several variants of JEVA will be presented. To further improve the performance of JEVA, we integrate JEVA with the list Viterbi algorithm (LVA) and propose the joint erasure marking and list Viterbi algorithm (JELVA). The JELVA provides two degrees of freedom in marking erasures and significantly improves the performance of JEVA with the cost of increased maximum num-ber of computed candidate codewords. To maintain the same maximum number of candidate codewords, we take the advantages of JEVA and LVA in different decoding stages and propose a switched JELVA, which is shown to be able to outperform both JEVA and LVA.

Finally, we apply the idea of JED to cognitive radio systems. Due to the bursty nature of interference in cognitive radio systems, the interference cannot be fully detected by pilots. To combat the residual undetected interference in the data packet, we apply a tailored joint erasure marking and decoding scheme for such systems. The proposed scheme is able to successfully detect the interfer-ence and achieve a performance close to that of the optimal decoder with the full interference knowledge including the frequency band and the power. With com-plexity reduction techniques, the proposed decoding scheme is only marginally more complex than the conventional schemes.