Source author record

Sofiene Affes

Sofiene Affes appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record

eess.SP Information Theory math.IT Applications Machine Learning math.OC

Catalog footprint

What is connected

4works

6topics

4close collaborators

Actions

Connect this record

Open graph Browse works

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

preprint2020arXiv

ANN-Based Detection in MIMO-OFDM Systems with Low-Resolution ADCs

In this paper, we propose a multi-layer artificial neural network (ANN) that is trained with the Levenberg-Marquardt algorithm for use in signal detection over multiple-input multiple-output orthogonal frequency-division multiplexing (MIMO-OFDM) systems, particularly those with low-resolution analog-to-digital converters (LR-ADCs). We consider a blind detection scheme where data symbol estimation is carried out without knowing the channel state information at the receiver (CSIR)---in contrast to classical algorithms. The main power of the proposed ANN-based detector (ANND) lies in its versatile use with any modulation scheme, blindly, yet without a change in its structure. We compare by simulations this new receiver with conventional ones, namely, the maximum likelihood (ML), minimum mean square error (MMSE), and zero-forcing (ZF), in terms of symbol error rate (SER) performance. Results suggest that ANND approaches ML at much lower complexity, outperforms ZF over the entire range of assessed signal-to-noise ratio (SNR) values, and so does it also, though, with the MMSE over different SNR ranges.

preprint2020arXiv

Coverage Analysis and Scaling Laws of Ultra-Dense Networks

In this paper, we develop an innovative approach to quantitatively characterize the performance of ultra-dense wireless networks in a plethora of propagation environments. The proposed framework has the potential of significantly simplifying the cumbersome procedure of analyzing the coverage probability and allowing the remarkable unification of single- and multi-antenna networks through compact representations. By harnessing this key feature, we develop a novel statistical machinery to study the scaling laws of wireless network densification considering general channel power distributions including the entire space of multipath and shadowing models as well as associated beamforming gain due to the use of multiple antenna. We further formulate the relationship between network density, antenna height, antenna array seize and carrier frequency showing how the coverage probability can be maintained with ultra-densification. From a system design perspective, we present a new innovative theoretical discovery stipulating that if multiple antenna BS are deployed and moved to higher frequencies, then monotonically increasing the coverage probability by means of ultra-densification is possible, and this without lowering the antenna height. Such findings are completely different from the conclusions in existing works, who suggest to lower the BS height as to leverage the potential of network densification. Simulation results substantiate performance trends leveraging network densification and antenna deployment and configuration against path loss models and signal-to-noise plus interference (SINR) thresholds.

preprint2014arXiv

Maximum Likelihood SNR Estimation of Linearly-Modulated Signals over Time-Varying Flat-Fading SIMO Channels

In this paper, we tackle for the first time the problem of maximum likelihood (ML) estimation of the signal-to-noise ratio (SNR) parameter over time-varying single-input multiple-output (SIMO) channels. Both the data-aided (DA) and the non-data-aided (NDA) schemes are investigated. Unlike classical techniques where the channel is assumed to be slowly time-varying and, therefore, considered as constant over the entire observation period, we address the more challenging problem of instantaneous (i.e., short-term or local) SNR estimation over fast time-varying channels. The channel variations are tracked locally using a polynomial-in-time expansion. First, we derive in closed-form expressions the DA ML estimator and its bias. The latter is subsequently subtracted in order to obtain a new unbiased DA estimator whose variance and the corresponding Cramér-Rao lower bound (CRLB) are also derived in closed form. Due to the extreme nonlinearity of the log-likelihood function (LLF) in the NDA case, we resort to the expectation-maximization (EM) technique to iteratively obtain the exact NDA ML SNR estimates within very few iterations. Most remarkably, the new EM-based NDA estimator is applicable to any linearly-modulated signal and provides sufficiently accurate soft estimates (i.e., soft detection) for each of the unknown transmitted symbols. Therefore, hard detection can be easily embedded in the iteration loop in order to improve its performance at low to moderate SNR levels. We show by extensive computer simulations that the new estimators are able to accurately estimate the instantaneous per-antenna SNRs as they coincide with the DA CRLB over a wide range of practical SNRs.

preprint2013arXiv

Quasi-convexity of the asymptotic channel MSE in regularized semi blind estimation

In this paper, the quasi-convexity of a sum of quadratic fractions in the form $\sum_{i=1}^n \frac{1+c_i x^2}{\left(1+d_ix\right)^2}$ is demonstrated where $c_i$ and $d_i$ are strictly positive scalars, when defined on the positive real axis $\mathbb{R}^{+}$. It will be shown that this quasi-convexity guarantees it has a unique local (and hence global) minimum. Indeed, this problem arises when considering the optimization of the weighting coefficient in regularized semi-blind channel identification problem, and more generally, is of interest in other contexts where we combine two different estimation criteria. Note that V. Buchoux {\it et.al} have noticed by simulations that the considered function has no local minima except its unique global minimum but this is the first time this result, as well as the quasi-convexity of the function is proved theoretically.