Probabilistic Equalization With a Smoothing Expectation Propagation Approach

Santos, Irene; Murillo-Fuentes, Juan José; Arias-de-Reyna, Eva; Olmos, Pablo M.

doi:10.1109/twc.2017.2672746

Cited by 7 publications

(30 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…where ppy k |s k q is given by the channel model in (1) and pps z1 k´1 |y 1:k´1 q denotes the marginal of pps k´1 |y 1:k´1 q over its first entry, i.e., over u k´L . Similarly, and in parallel, a backward procedure can be run following the same formulation explained above by just left-right flipping the channel, received and transmitted vectors [11]. As a result, the distribution pps k |y k:N`L q is estimated.…”

Section: From the Bcjr To The Lmmsementioning

confidence: 99%

“…Also, this equalizer is just designed for turbo equalization, boiling down to the LMMSE for standalone equalization. In [11] the authors proposed a self-iterated EP equalizer, the smoothing expectation propagation (SEP), that improves the performance either as standalone or turbo equalization. The EP was introduced to improve the estimation of the posteriors in the forward and backward approaches.…”

Section: Introductionmentioning

confidence: 99%

“…In this paper we propose a new Kalman smoothing EP equalizer where we have introduced the following improvements. First, the EP is introduced at the smoothing step, improving the convergence in comparison to [11]. Second, the true prior used in the moment matching procedure of the EP algorithm is set to a non-uniform distribution provided by the channel decoder [23].…”

Section: Introductionmentioning

confidence: 99%

“…Second, the true prior used in the moment matching procedure of the EP algorithm is set to a non-uniform distribution provided by the channel decoder [23]. This is a further difference with respect to the SEP approach in [11], where uniform priors were used even after the turbo procedure. Third, a better control of the minimum allowed variance at the moment matching is introduced.…”

Section: Introductionmentioning

confidence: 99%

“…Finally, we review the forgetting factor used at the damping procedure. As a result, better gains in terms of bit error rate (BER) are achieved, the convergence in turbo equalization is more robust compared to the one of the equalizers in [11], [24] and the computational complexity yields quadratic in the channel length, i.e., the one of the Kalman smoothing solution.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Channel Equalization With Expectation Propagation at Smoothing Level

Santos

Murillo-Fuentes

Aradillas

et al. 2020

IEEE Trans. Commun.

Self Cite

View full text Add to dashboard Cite

In this paper we propose a smoothing turbo equalizer based on the expectation propagation (EP) algorithm with quite improved performance compared to the Kalman smoother, at similar complexity. In scenarios where high-order modulations or/and large memory channels are employed, the optimal BCJR algorithm is computationally unfeasible. In this situation, lowcost but suboptimal solutions, such as the linear minimum mean square error (LMMSE), are commonly used. Recently, EP has been proposed as a tool to improve the Kalman smoothing performance. In this paper we review these solutions to apply the EP at the smoothing level, rather than at the forward and backwards stages. Also, we better exploit the information coming from the channel decoder in the turbo equalization schemes. With these improvements we reduce the computational complexity, speed up convergence and outperform previous approaches. We included some simulation results to show the robust behavior of the proposed method regardless of the scenario, and its improvement in terms of performance in comparison with other EP-based solutions in the literature.

show abstract

Section: From the Bcjr To The Lmmsementioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%