Pilot Pattern Design for Deep Learning-Based Channel Estimation in OFDM Systems

Abstract

In this paper, we present a downlink pilot design scheme for Deep Learning (DL) based channel estimation (ChannelNet) in orthogonal frequency-division multiplexing (OFDM) systems. Specifically, in the proposed scheme, a feature selection method named Concrete Autoencoder (ConcreteAE) is used to find the most informative locations for pilot transmission. This autoencoder consists of a concrete layer as the encoder and a multilayer perceptron (MLP) as the decoder. During the training, the concrete layer selects the most informative pilot locations, and the decoder reconstructs an approximate estimation of the channel. Eventually, the ChannelNet is trained on the output of the ConcreteAE aiming to reconstruct the ideal channel response. The estimation error results show that this approach outperforms the previously presented ChannelNet with a uniformly distributed pilot pattern, and its performance is comparable to the minimum mean square error (MMSE).

A survey of applied machine learning techniques for optical orthogonal frequency division multiplexing based networks

“…6) Finally, deep learning is being an attractive technique that take more attention from researcher community implemented for wireless and optical OFDM-based systems in various applications like channel estimation [107,108], impairment compensation [109] and signal modulation identification [110].…”

“…ML can be viewed as a promising technique for routing and wavelength allocation in WDM based satellite optical networks 119 . Due to the satellite mobility around orbits nature leading to apparition of various undesirable effects such as transmission delay and Doppler shift, reinforcement learning can be an effective solution for learning from the environment to produce a real‐time prediction for routing and wavelength selection. Deep learning (ie, DNN) is an attractive technique that has taken more attention from the researcher community for wireless and O‐OFDM‐based systems used in various applications like channel estimation, 120,121 impairment compensation, 122 and signal modulation identification 123 . Table 5 presents the simple/hybrid deep learning algorithms implemented in image classification, IoT and network intrusion detection system (IDS) applications that can be a potential candidate for O‐OFDM networks as a function of the dataset used, the performance measured metric (ie, accuracy) and the optimized setting. It envisioned ML techniques to be effective and valuable in 5G new radio (NR) based systems as well.…”

An Asymptotically Optimal Approximation of the Conditional Mean Channel Estimator based on Gaussian Mixture Models

“…The concept of a concrete autoencoder (CAE) was introduced in [40] and adapted for wideband channel estimation (cf. Section II-C) in [41]. The CAE is an autoencoder where the encoder is replaced by a concrete selection layer that selects the N p ≪ N c N t most informative features of the N c N tdimensional input.…”

“…Hence, a new CAE needs to be trained for every different SNR. In our simulations, in contrast to [41], no further denoising networks are applied after the CAE.…”

Joint Channel Estimation and Mixed-ADCs Allocation for Massive MIMO via Deep Learning

“…Recently, numerous works have shown the successful applications of deep learning (DL) in communication systems, which includes channel estimation [20]- [23], pilot design [24]- [27], CSI feedback [28]- [31], data detection [32]- [35], et al DL algorithms exhibits superior performance in tackling the one-bit channel estimation problem that the traditional methods are unable to handle well. Specifically, the authors in [22] have devised a novel DL-based architectures for the one-bit quantized orthogonal frequency division multiplexing (OFDM) receiver to deal with the channel estimation and data detection problem.…”