The objective of this thesis is to improve the performance of the conventional turbo decoders by placing them inside an iterative loop which uses optimization techniques to correct codewords which fail to converge on the initial decoding pass. The increase in bit error and code word error rate performance of the decoder can be achieved by adding small noise increments to “mutate” the received signal in the direction of convergence. Optimization algorithm techniques are used to increase the bit error and code word error rate performance of the decoder.
A conventional MAX-log-MAP algorithm is used to decode most codewords. Simple hill climbing and random restart hill climbing algorithms are used when decoder fails to converge on the initial decoding attempt.
In this thesis the conventional turbo decoder has been studied via computer simulations by applying optimization algorithm techniques and by introducing small artificial noise increments to the input signal. Improvements of an order of magnitude in bit error rate are achieved.