Tree Search Detection Algorithms for Equalizing Single-Carrier Block Transmission Signals and Separating Spatially Multiplexed Signals

Abstract

In this dissertation we propose efficient tree search detection algorithms to combat two types of interference that commonly arises in wireless communication systems: (i) inter-symbol interference (ISI) in the single-carrier block transmission (SCBT) systems in frequency-selective channels and (ii) inter-antenna interference (IAI) in spatial multiplexing (SM) multiple-input multiple-output (MIMO) systems For SCBT systems a conventional low-complexity approach to ISI mitigation is to use frequency-domain equalization (FDE) However FDE in the literature is limited to symbol-by-symbol detection only such as the frequency-domain linear equalizer iterative block decision feedback equalizer (IBDFE) as well as the hybrid decision feedback equalizer (HDFE) that combines a frequency-domain feedforward filter and a time-domain feedback filter Although with a low complexity the performance of these detectors considerably lag behind that of the optimal maximum-likelihood sequence detector (MLSD) To improve the performance of symbol-by-symbol detectors while preserving the implementation advantages of FDE we propose in this dissertation two hybrid-domain sequence detectors The first one referred to as the FDF-M algorithm combines FD prefiltering with time-domain tree search using the M-algorithm The proposed FDF-M algorithm achieves a detection performance that is very close to that of the conventional time-domain sequence detectors (which employ the QRD-M or sphere decoding algorithm) but with a much lower complexity Due to running the M-algorithm however the complexity of FDF-M is still quite high (with respect to that of HDFE) for high-order modulation and the tree search stage does not exhibit a high degree of parallelism To address these issues we propose the second hybrid-domain sequence detector referred to as the parallel HDFE (P-HDFE) algorithm It adaptively operates based on the reliability of the decision variable as an ordinary HDFE or a tree search detector constructed by multiple HDFEs that run in parallel Our study shows that at moderate to high signal-to-noise ratios (SNRs) P-HDFE significantly outperforms HDFE with little increase in complexity We analyze the symbol-error rate of P-HDFE at high SNRs for the case of quaternary phase-shift keying and static ISI channels accounting for error propagation and residual ISI Simulations demonstrate the accuracy of the analysis For MIMO systems hybrid tree search algorithms for maximum likelihood symbol detection are described Essentially the search tree is iteratively expanded in the breadth-first (BF) manner until the probability that the current most likely path is correct exceeds the specified threshold at which point the depth-first (DF) stage is initiated to traverse the rest of the tree In contrast to the sphere decoding algorithm (SDA) which starts off with the DF search the proposed algorithms use the BF stage to enhance the accuracy of the initial DF search direction by exploiting the diversity inherent in the SM scheme Simulation results demonstrate that with a moderate increase in the memory requirement the proposed algorithms achieve a significantly lower complexity than the SDA in many scenarios

Date of Award	2017 Jan 20
Original language	English
Supervisor	Kuei-Chiang Lai (Supervisor)