A low-complexity Max-Log-MAP detector
IEEE Transactions on Communications
On the relation of MIMO APP detection and SIMO maximum ratio combining
GLOBECOM'09 Proceedings of the 28th IEEE conference on Global telecommunications
Reliability-based hybrid MMSE/subspace-max-log-APP MIMO detector
IEEE Communications Letters
| Hi-index | 0.00 |
The Soft-Output M-Algorithm (SOMA) is a reduced-complexity trellis decoder based on a sequential decoding technique known as the M-algorithm. Instead of extending all survivors from one trellis depth to the next, it extends only from the best M. The remaining survivors are terminated. The novelty of the SOMA is the use of terminated paths to obtain reliable soft-information. Soft-information is extracted from terminated paths through a simple update-and-discard procedure. It is found that using the M best fully-extended survivors alone is inadequate to deliver soft-information due to their similarity. On the other hand, terminated paths of a pruned trellis carry a significant amount of soft-information which should not be ignored for reliable bit-detection. The SOMA is particularly useful in reducing the complexity of an iterative receiver. Applications include turbo equalization of single-input single-output Inter-Symbol Interference (ISI) channels and Multiple-Input Multiple-Output (MIMO) frequency selective fading channels, iterative decoding of MIMO flat fading channels, and multi-user detection and equalization in coded Direct-Sequence Code-Division Multiple-Access (DS-CDMA) systems. For turbo equalization of a 16-tap ISI channel with convolutional coding, nearoptimal bit-error-rate is achieved by retaining only 128 (out of 32,768) states. The SOMA is also applicable to reduce the complexity of tree decoding. For a coded MIMO system with 8 antennae transmitting 64-QAM symbols over a flat fading channel, the respective code-tree with 648 hypotheses can be decoded with the SOMA by retaining only 32 paths per tree depth. The complexity of multi-user detection for coded DS-CDMA systems for flat-fading channels can also be significantly reduced using the SOMA. High-order trellises with millions of states can also be decoded with modest complexity using the SOMA. The trellis is first expanded into an equivalent trellis/tree structure. Then, the M-algorithm is applied twice, once to reduce the number of states, and the other to reduce the number of branches emanating from each state. The proposed method gives promising results in equalizing a high-order modulated MIMO system with bit-interleaved coded modulation undergoing frequency selective fading. The method is also applicable to reduce the complexity of multi-user equalization of a coded DS-CDMA system undergoing frequency selective fading.