Title | Sequence detection and adaptive channel estimation for ISI channels under class-a impulsive noise |
Publication Type | Journal Article |
Year of Publication | 2004 |
Authors | Schober, R., and L. Lampe |
Journal | Communications, IEEE Transactions on |
Volume | 52 |
Pagination | 1523 - 1531 |
Date Published | sep. |
ISSN | 0090-6778 |
Keywords | adaptive channel estimation scheme, AWGN channels, channel equalization, channel estimation, class-A impulsive noise, computational complexity, Entropy, error probability, error statistics, error variance, fading channels, frequency-selective channels, Gaussian noise, impulse noise, intersymbol interference, ISI channels, least mean squares methods, least-mean entropy algorithm, matched filter, matched filters, maximum likelihood estimation, maximum-likelihood sequence detection, recursive estimation, recursive least-entropy algorithm, signal detection, suboptimum sequence detection scheme |
Abstract | In this paper, sequence detection and channel estimation for frequency-selective, intersymbol interference (ISI)-producing channels under Class-A impulsive noise are considered. We introduce a novel suboptimum sequence detection (SSD) scheme and show that although SSD employs a simplified metric, it achieves practically the same performance as maximum-likelihood sequence detection (MLSD). For both SSD and MLSD, a lower bound on the achievable performance is derived, which is similar to the classical matched-filter bound for frequency-selective (fading) channels under Gaussian noise. For channel estimation, we adopt a minimum entropy criterion and derive efficient least-mean-entropy and recursive least-entropy algorithms. For both adaptive algorithms, we analyze the steady-state channel-estimation error variance. Theoretical considerations and simulation results show that in Class-A impulsive noise, the proposed sequence detection and adaptive channel-estimation schemes yield significant performance gains over their respective conventional counterparts (designed for Gaussian noise). Although the novel algorithms require knowledge of the Class-A noise-model parameters, their computational complexity is comparable to that of the corresponding conventional algorithms. |
URL | http://dx.doi.org/10.1109/TCOMM.2004.833197 |
DOI | 10.1109/TCOMM.2004.833197 |