Abstract <p>Effective Electroencephalogram (EEG) signal processing necessitates the mitigation of physiological artifacts. While deep learning frameworks have demonstrated superior performance over traditional methods for this task, their high complexity and computational demands hinder deployment on resource-constrained platforms. In this work, denoising network called EEGPARnet is proposed to address this limitation. The proposed architecture integrates transformer encoders equipped with temporal and spectral attention modules and a Gated Recurrent Unit (GRU)-based decoder. This fusion enables the model to learn time-frequency long-range similarities, facilitating efficient feature extraction and a reduced number of trainable parameters. Experimental validation of the proposed model on the EEGDenoiseNet dataset revealed an average temporal relative root mean square error (<InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(RRMSE_t\)</EquationSource> </InlineEquation>) of 0.289, spectral relative root mean square error (<InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(RRMSE_s\)</EquationSource> </InlineEquation>) of 0.312, and a correlation coefficient (CC) of 0.942 for ocular artifact removal. For muscular artifact removal, the proposed method achieved competitive results against state-of-the-art techniques, with mean <InlineEquation ID="IEq3"> <EquationSource Format="TEX">\(RRMSE_t\)</EquationSource> </InlineEquation>, <InlineEquation ID="IEq4"> <EquationSource Format="TEX">\(RRMSE_s\)</EquationSource> </InlineEquation>, and CC values of 0.458, 0.428, and 0.855, respectively. Compared to state-of-the-art model, the proposed EEGPARnet demonstrated a significant reductions in computational complexity with <InlineEquation ID="IEq5"> <EquationSource Format="TEX">\({\textbf {84}}\times\)</EquationSource> </InlineEquation> fewer trainable parameters, <InlineEquation ID="IEq6"> <EquationSource Format="TEX">\({\textbf {30}}\times\)</EquationSource> </InlineEquation> less FLOPS, and <InlineEquation ID="IEq7"> <EquationSource Format="TEX">\({\textbf {160}}\times\)</EquationSource> </InlineEquation> smaller storage, making it a step closer towards deployment on resource-constrained devices for real-time EEG denoising without compromising performance.</p> Graphical abstract <p></p>

EEGPARnet: time-frequency attention transformer encoder and GRU decoder for removal of ocular and muscular artifacts from EEG signals

期刊论文
发表日期：

2025年12月28日

查看全文

Kiyam Babloo Singh,
Aheibam Dinamani Singh,
Merin Loukrakpam

摘要

Abstract

Effective Electroencephalogram (EEG) signal processing necessitates the mitigation of physiological artifacts. While deep learning frameworks have demonstrated superior performance over traditional methods for this task, their high complexity and computational demands hinder deployment on resource-constrained platforms. In this work, denoising network called EEGPARnet is proposed to address this limitation. The proposed architecture integrates transformer encoders equipped with temporal and spectral attention modules and a Gated Recurrent Unit (GRU)-based decoder. This fusion enables the model to learn time-frequency long-range similarities, facilitating efficient feature extraction and a reduced number of trainable parameters. Experimental validation of the proposed model on the EEGDenoiseNet dataset revealed an average temporal relative root mean square error ( \(RRMSE_t\) ) of 0.289, spectral relative root mean square error ( \(RRMSE_s\) ) of 0.312, and a correlation coefficient (CC) of 0.942 for ocular artifact removal. For muscular artifact removal, the proposed method achieved competitive results against state-of-the-art techniques, with mean \(RRMSE_t\) , \(RRMSE_s\) , and CC values of 0.458, 0.428, and 0.855, respectively. Compared to state-of-the-art model, the proposed EEGPARnet demonstrated a significant reductions in computational complexity with \({\textbf {84}}\times\) fewer trainable parameters, \({\textbf {30}}\times\) less FLOPS, and \({\textbf {160}}\times\) smaller storage, making it a step closer towards deployment on resource-constrained devices for real-time EEG denoising without compromising performance.

Graphical abstract