[Paper Review] On the Log-Likelihood Ratio Evaluation of CWCU Linear and Widely Linear MMSE Data Estimators

Oliver Lang, Mario Huemer+1 more|arXiv (Cornell University)|Jul 8, 2016

Advanced Wireless Communication Techniques9 references3 citations

TL;DR

This paper proves that component-wise conditionally unbiased (CWCU) linear and widely linear MMSE estimators produce identical log-likelihood ratios (LLRs) as their standard LMMSE and WLMMSE counterparts for proper (e.g., QPSK) and improper (e.g., 8-QAM) constellations, respectively. Despite higher BMSE, the CWCU estimators enable lower-complexity LLR computation without BER performance loss, offering a practical advantage in soft decoding systems.

ABSTRACT

In soft decoding of data bits, the log-likelihood ratios are evaluated from the estimated data symbols. For proper constellation diagrams such as QPSK or 16-QAM, these data symbols are often estimated using the linear minimum mean square error (LMMSE) estimator. The LMMSE estimator only fulfills the weak Bayesian unbiasedness constraint. Recently, estimators fulfilling the more stringent component-wise conditionally unbiased (CWCU) constraints have been investigated, such as the CWCU LMMSE estimator. In this paper, we prove that the CWCU LMMSE estimates result in the very same log-likelihood ratios as the LMMSE estimates. For improper constellation diagrams such as 8-QAM, widely linear estimators are used. For this case, we show that the widely linear versions of the LMMSE estimator and the CWCU LMMSE estimator also yield identical log-likelihood ratios. Finally, we give a simulation example which illustrates a number of interesting properties of the discussed widely linear estimators.

Motivation & Objective

• To analyze the relationship between component-wise conditionally unbiased (CWCU) estimators and standard MMSE estimators in terms of log-likelihood ratio (LLR) output.
• To determine whether the CWCU LMMSE estimator produces the same LLRs as the conventional LMMSE estimator for proper constellations like QPSK.
• To investigate whether the CWCU widely linear MMSE (WLMMSE) estimator yields identical LLRs to the standard WLMMSE estimator for improper constellations such as 8-QAM.
• To demonstrate that the CWCU estimators offer a computational advantage in LLR evaluation without sacrificing bit error rate (BER) performance.
• To validate the theoretical findings through a simulation example using UW-OFDM with 8-QAM over AWGN and frequency-selective channels.

Proposed method

• Theoretical derivation proves that the conditional mean of the CWCU LMMSE estimator satisfies the same unbiasedness condition per symbol as the standard LMMSE estimator under proper constellation assumptions.
• For improper constellations like 8-QAM, the widely linear MMSE (WLMMSE) estimator is used as the baseline, and the CWCU WLMMSE estimator is derived to satisfy component-wise conditional unbiasedness.
• The log-likelihood ratio (LLR) is computed using the conditional probability density function (PDF) of the estimated symbol given the transmitted symbol, assuming Gaussian approximation via the central limit theorem.
• The analysis compares the LLR expressions derived from the standard and CWCU estimators, showing equivalence under the same model assumptions.
• A simulation framework using UW-OFDM with 36-length data vectors and 52-frequency-domain samples is implemented to validate the theoretical results over AWGN and frequency-selective channels.
• The statistical properties of the estimates—specifically, the properness of the conditional distribution—are evaluated by analyzing the covariance matrix of the estimates given a transmitted symbol.

Experimental results

Research questions

• RQ1 Do the log-likelihood ratios (LLRs) computed from the CWCU LMMSE estimator match those from the standard LMMSE estimator for proper constellations such as QPSK?
• RQ2 Does the CWCU widely linear MMSE (WLMMSE) estimator produce the same LLRs as the standard WLMMSE estimator for improper constellations like 8-QAM?
• RQ3 Can the CWCU estimators reduce the computational complexity of LLR evaluation without degrading BER performance?
• RQ4 Are the conditional estimates from the CWCU WLMMSE estimator properly distributed when the transmitted symbols are from an improper constellation like 8-QAM?
• RQ5 How do the statistical properties (e.g., mean and covariance) of the CWCU WLMMSE estimates compare to those of the standard WLMMSE estimates in frequency-selective fading channels?

Key findings

• The CWCU LMMSE estimator produces identical log-likelihood ratios (LLRs) as the standard LMMSE estimator for proper constellations such as QPSK, despite differing in mean square error performance.
• For improper constellations like 8-QAM, the CWCU WLMMSE estimator yields the same LLRs as the standard WLMMSE estimator, ensuring equivalent bit error rate (BER) performance.
• The CWCU WLMMSE estimates are properly distributed given a transmitted symbol, allowing the use of the simpler proper complex Gaussian PDF for LLR computation.
• This proper distribution enables a complexity reduction in LLR evaluation, as the general bivariate complex Gaussian PDF is no longer required.
• In simulations over UW-OFDM with 8-QAM, the off-diagonal elements of the conditional covariance matrix of the CWCU WLMMSE estimates were smaller than the diagonal elements by a factor of at least 1000, confirming near-proper behavior.
• The BER performance of the CWCU WLMMSE estimator was indistinguishable from that of the standard WLMMSE estimator, even in frequency-selective fading channels, validating the theoretical equivalence of LLRs.