<p>In this paper, we propose a two-stage, deep learning–based approach to enhance channel estimation in narrow-band power line communication (NB-PLC) OFDM systems. The method is designed for channels which are time- and frequency-selective and operate in the presence of impulsive noise. The first stage involves a convolutional neural network (CNN) to denoise the received OFDM signal in the time domain, while the second stage is based on a deep neural network (DNN) which estimates the NB-PLC channel coefficients in the frequency domain from the denoised signal. Simulation results demonstrate that the proposed method provides better performance in terms of mean square error (MSE) and bit error rate (BER) compared to the conventional linear minimum mean square error (LMMSE) and least squares (LS) estimators. For a signal-to-noise ratio (SNR) of 15 dB, the proposed CNN-DNN model achieves an average MSE gain of approximately (<InlineEquation ID="IEq1"> <EquationSource Format="TEX">$7 \times 10^{-5}$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>7</mn> <mo>×</mo> <msup> <mn>10</mn> <mrow> <mo>-</mo> <mn>5</mn> </mrow> </msup> </mrow> </math></EquationSource> </InlineEquation>) and (<InlineEquation ID="IEq2"> <EquationSource Format="TEX">$1 \times 10^{-4}$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>1</mn> <mo>×</mo> <msup> <mn>10</mn> <mrow> <mo>-</mo> <mn>4</mn> </mrow> </msup> </mrow> </math></EquationSource> </InlineEquation>) compared to the LMMSE and LS techniques, respectively. Alternatively, for a BER of (<InlineEquation ID="IEq3"> <EquationSource Format="TEX">$1 \times 10^{-1}$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>1</mn> <mo>×</mo> <msup> <mn>10</mn> <mrow> <mo>-</mo> <mn>1</mn> </mrow> </msup> </mrow> </math></EquationSource> </InlineEquation>), SNR gains of 1.5 dB and 10 dB are obtained with our model compared to LMMSE and LS, respectively. Additionally, the proposed approach demonstrates reduced computational complexity relative to the LMMSE method.</p>

Enhanced deep learning–based channel estimator for OFDM narrowband powerline communication systems

期刊论文
发表日期：

2026年3月4日

查看全文

Wided Belhaj Sghaier,
Héla Gassara,
Fatma Rouissi,
Fethi Tlili

摘要

In this paper, we propose a two-stage, deep learning–based approach to enhance channel estimation in narrow-band power line communication (NB-PLC) OFDM systems. The method is designed for channels which are time- and frequency-selective and operate in the presence of impulsive noise. The first stage involves a convolutional neural network (CNN) to denoise the received OFDM signal in the time domain, while the second stage is based on a deep neural network (DNN) which estimates the NB-PLC channel coefficients in the frequency domain from the denoised signal. Simulation results demonstrate that the proposed method provides better performance in terms of mean square error (MSE) and bit error rate (BER) compared to the conventional linear minimum mean square error (LMMSE) and least squares (LS) estimators. For a signal-to-noise ratio (SNR) of 15 dB, the proposed CNN-DNN model achieves an average MSE gain of approximately ( $7 \times 10^{-5}$ $7 \times 10^{- 5}$ ) and ( $1 \times 10^{-4}$ $1 \times 10^{- 4}$ ) compared to the LMMSE and LS techniques, respectively. Alternatively, for a BER of ( $1 \times 10^{-1}$ $1 \times 10^{- 1}$ ), SNR gains of 1.5 dB and 10 dB are obtained with our model compared to LMMSE and LS, respectively. Additionally, the proposed approach demonstrates reduced computational complexity relative to the LMMSE method.