We will investigate a general class of concatenated codes and coded-modulation, by building upon our recent invention. Our concatenated scheme is more general than the conventional concatenated codes in that a "coder" can be a channel with memory (e.g., a partial-response channel, a channel with intersymbol interference and/or multipath delays, a channel with burst noise), or a modulator with some constraint or memory (e.g., continuous phase modulation (CPM), trellis-coded modulation (TCM)). We can also concatenate more than two coders. In our proposed system an "ambiguity zone detector (AZD)" plays a critical role. The AZD introduces generalized "erasure" symbols, and is found to be a most efficient way of exploiting soft-decision outputs. We also make use of a "permutation" and "iterative decoding", as has been done in Turbo codes. But unlike the Turbo codes and such other known schemes as product codes, the permutation size required for our system needs not be so large. The main advantage that differentiates our approach from prior arts is its extremely simple decoding algorithm. Hence, the decoder is easily implementable and the decoding delay is kept minimal. The performance of our system, measured in terms of residual errors/erasures, is extremely good. For example, our initial result obtained for a system which adopts a Hamming code and a partial-response channel G(D)=1+D demonstrates that our decoding algorithm converges within several iterations, and produces nearly error-free outputs even when more than 50% of the AZD outputs are labeled "erasures". In this proposed study we intend to perform the following research tasks: 1. Simulation of specific application examples (e.g., partial-response channels, wireless channels with CPM) 2. Determination of good permutations and effective decoders (e.g., optimal selections of the ambiguity zones) 3. Performance analysis of the proposed codes, and coded modulation schemes (e.g., bounds and approximation for t he decoding error probability) 4. Construction of powerful codes, using a more general concatenation topology (e.g. a parallel-serial concatenation), and the corresponding decoding procedures. We also expect that insights to be gained in the present study should lead to a better understanding of the Turbo codes, and vice versa.
