Soft information processing in BICM and relay systems : complexity reduction and code design

Abstract

Digital communication systems are today close to achieving the limits that were predicted by Claude Shannon only a few decades ago.<br />Core concepts of such high-performance systems are soft information and iterative processing: receiver blocks like channel estimator, equalizer, demodulator and channel decoder repeatedly compute and exchange log-likelihood ratios (LLRs) of the transmitted bits before the final decision is made. Another recent paradigm in wireless networks is cooperative communication: nodes in a network exploit the broadcast nature of the wireless channel and help each other to receive their intended information, thereby realizing gains in communication reliability, power efficiency, and transmission rates. Cooperation is especially important in wireless sensor networks, where a large number of nodes needs to transmit possibly correlated data over long distances.<br />Besides the striking performance benefits, systems with iterative processing based on soft informa- tion are more complex and more difficult to implement. Complexity reduction is therefore a vital issue.<br />In the first part of this thesis, we design and analyze reduced-complexity receivers for bit-interleaved coded modulation (BICM). LLR clipping is analyzed using concepts of information geometry and the system capacity is evaluated numerically. We also propose a selective LLR update scheme for an iterative BICM receiver, which reduces computational complexity by omitting computations that have negligible impact on the performance.<br />In the second part, a cooperative transmission scheme for the three-node relay channel based on quantized soft information is proposed. The relay demodulates the received data and quantizes the resulting LLRs, before transmitting them to the destination. Numerical simulations show that the proposed scheme delivers a performance comparable to amplify-and-forward, while having the advantage of an all-digital implementation.<br />Finally, we consider a relay system with two correlated sources. For the uncoded case, we derive analytical bounds on the error probability, both for optimum and sub-optimum detection. Then we propose a distributed coding scheme based on spatially-coupled low-density parity-check codes, which encompasses joint source, network, and channel coding. We derive theoretical limits of the achievable system rate and prove that our scheme achieves these limits for independent sources and symmetric link qualities. Based on numerical simulations of density evolution, we conjecture that our codes achieve the limits also for correlated sources and asymmetric channels.