This event has been postponed and will no longer take place on 25 March. We are hoping to rearrange another date soon.
This talk will present a family of novel, non-statistical, Euclidean-distance soft-input, soft-output (SISO) decoding algorithms.
These algorithms are universal because they can be applied to error-correcting codes with any combination of parameters (length, distance, rate, etc). They are also universal because their processing structure can be based on any existing (and future) decoding structure. The novelty of these algorithms is that their metric is squared Euclidean distance, which is a non-statistical metric in contrast to the statistical metrics used in almost all existing decoders.
Results will be presented of simulations comparing the performance of LDPC, convolutional and polar codes, respectively decoded using the iterative sum-product and BCJR structures and the non-iterative successive cancellation structure, over AWGN, fading and impulse-noise channels, with either the novel non-statistical metric or the usual statistical metric in each case.
In all cases, the non-statistical decoding algorithms have performed as well or slightly better than the existing statistical algorithms, providing near-optimum results. In addition, an 8% reduction in processing complexity is achieved over the AWGN channel; in all cases equipment to measure or estimate the channel statistics is not required. So there are significant advantages in using the non-statistical algorithms.