Radial Basis Function Assisted Turbo Equalization

Abstract

This paper presents a turbo equalization (TEQ) scheme, which employs a radial basis function (RBF)-based equalizer instead of the conventional trellis-based equalizer of Douillard et al. Structural, computational complexity, and performance comparisons of the RBF-based and trellis-based TEQs are provided. The decision feedback-assisted RBF TEQ is capable of attaining a similar performance to the logarithmic maximum $a posteriori$ scheme in the context of both binary phase-shift keying (BPSK) and quaternary phase-shift keying (QPSK) modulation, while achieving a factor 2.5 and 3 lower computational complexity, respectively. However, there is a 2.5-dB performance loss in the context of 16 quadrature amplitude modulation (QAM), which suffers more dramatically from the phenomenon of erroneous decision-feedback effects. A novel element of our design, in order to further reduce the computational complexity of the RBF TEQ, is that symbol equalizations are invoked at current iterations only if the decoded symbol has a high error probability. This techniques provides 37% and 54% computational complexity reduction compared to the full-complexity RBF TEQ for the BPSK RBF TEQ and 16QAM RBF TEQ, respectively, with little performance degradation, when communicating over dispersive Rayleigh fading channels. Index Terms—Decision-feedback equalizer (DFE), Jacobian logarithm, neural network, radial basis function (RBF), turbo coding, turbo equalization (TEQ).

More information

Identifiers

ORCID for L. Hanzo: orcid.org/0000-0002-2636-5214

Catalogue record

Export record