Efficient complexity reduction methods for short-frame iterative decoding

Turbo coding has already become a prominent error correction technique for wireless communications, especially after its adaptation in the cdma2000 and W-CDMA third generation mobile radio standards. It is well-known that the complexity of iterative decoding of Turbo codes can be greatly reduced by applying an appropriate stopping criterion, instead of performing a fixed number of iterations. Existing stopping criteria are mostly applied after every iteration. While these criteria are very efficient for large frame sizes, they are not necessarily the best for small frame sizes. In this thesis we investigate how conventional stopping criteria can be enhanced for low-complexity iterative decoding of short-frame Turbo codes with two or more component codes. It is shown that decoding complexity can be reduced by adapting conventional iteration-based stopping criteria to component-based variants with a suitable step size, which is a parameter introduced for complexity minimization in multiple-component Turbo codes of short frame sizes. In addition, further complexity reduction can be obtained by properly strengthened the stopping criterion in use. Our simulation results demonstrate that an impressive complexity saving of 38.8% for a rate-1/2 Turbo codes with frame size 84 bits over an AWGN channel can be achieved by applying the proposed methods to the simple hard-decision-aided (HDA) stopping criterion. Next, the proposed complexity reduction methods are extended for the Chase-algorithm-based iterative decoding of BCH product codes (PC) and are found to result in even more attractive saving for short-frame product codes. In combination with the use of stopping criteria, we also propose methods for reducing the number of test patterns used in the Chase algorithm, resulting in very low-complexity iterative decoding schemes. Simulation results show that a 50% saving in the decoding complexity can be achieved for some example product codes. Finally, low-complexity iterative decoders that are robust against estimation error in signal-to-noise-ratio (SNR) are designed. We propose a simple "switch-ing" iterative decoder, whose component decoders may switch from Log-MAP to MAX-Log-MAP decoders in a carefully designed manner. Simulation results show that a properly designed switching decoder is superior to both the conventional Log-MAP-based or MAX-Log-MAP-based iterative decoders in the presence of severe SNR estimation error