---
title: 'EEGDenoiseNet: EEG Denoising Benchmark'
url: https://www.emergentmind.com/topics/eegdenoisenet
type: topic
---

# EEGDenoiseNet: EEG Denoising Benchmark

Searching arXiv for EEGDenoiseNet and closely related benchmark papers to ground the article in current literature.
EEGDenoiseNet is a benchmark dataset and evaluation framework for supervised, end-to-end EEG denoising, introduced to provide standardized clean EEG targets, isolated artifact segments, and common benchmarks for deep-learning-based artifact removal [2009.11662]. It was created to address the absence of an open dataset explicitly structured for denoising research, where contaminated EEG can be synthesized with known ground-truth clean EEG, allowing fair comparison across models and controlled study of denoising under ocular and myogenic contamination [2009.11662]. In its original form, EEGDenoiseNet contains curated single-channel 2-second segments of clean EEG, ocular artifact, and muscular artifact, and has subsequently become a reference benchmark for studies on ocular artifact removal, EMG suppression, frequency-conditioned denoising, Transformer- and RetNet-based sequence models, GAN-based denoising, and capacity-controlled benchmark analysis [2009.11662], [2310.17335], [2109.04235], [2404.15289], [2411.13288], [2606.08594].

## 1. Dataset definition and construction

EEGDenoiseNet was introduced as a publicly available benchmark dataset specifically for deep-learning EEG denoising, rather than for general EEG analysis or BCI tasks [2009.11662]. Its defining feature is that it separates the data needed for supervised denoising into three curated pools: **4514 clean EEG segments**, **3400 ocular artifact segments**, and **5598 myogenic artifact segments** [2009.11662]. Each segment is **2 seconds long**, and the dataset is distributed in both **MATLAB `.mat`** and **Python NumPy `.npy`** formats [2009.11662].

The clean EEG pool was derived from the motor imagery dataset of Cho et al. using **52 participants**, **64-channel EEG**, and a **512 Hz** sampling rate; those recordings were processed with **1–80 Hz** band-pass filtering, a notch filter at the local powerline frequency, resampling to **256 Hz**, and artifact attenuation using **ICLabel**, followed by segmentation into **2 s** one-dimensional segments, standardization, and expert visual inspection [2009.11662]. The ocular artifact pool was extracted from open datasets containing horizontal and vertical EOG channels, with preprocessing consisting of **0.3–10 Hz** band-pass filtering, resampling to **256 Hz**, segmentation into **2 s** one-dimensional segments, standardization, and expert inspection [2009.11662]. The myogenic artifact pool was derived from facial EMG recordings, processed with **1–120 Hz** band-pass filtering, a notch filter at powerline frequency, resampling to **512 Hz**, segmentation into **2 s** segments, standardization, and expert visual inspection [2009.11662].

The dataset is intentionally **single-channel**, which the original paper presents as a design choice to improve universality across EEG systems rather than tie the benchmark to a fixed multichannel montage [2009.11662]. The 2-second segment length was chosen because it is long enough to preserve temporal and spectral characteristics of EEG, EOG, and EMG, yet short enough that artifact-free clean EEG segments can realistically be curated [2009.11662].

## 2. Semi-synthetic contamination protocol

EEGDenoiseNet is fundamentally a **semi-synthetic supervised denoising** benchmark. The released dataset does not store noisy-clean pairs directly; instead, it provides curated clean EEG and artifact segments from which contaminated EEG can be synthesized by linear mixing [2009.11662]. The canonical contamination model is

$$
y = x + \lambda \cdot n
$$

where \(y\) is the contaminated EEG, \(x\) is the clean EEG target, \(n\) is an artifact segment, and \(\lambda\) controls contamination strength and thus the signal-to-noise ratio [2009.11662]. The signal-to-noise ratio is defined through the RMS ratio between clean EEG and scaled artifact, with the RMS itself defined as

$$
RMS(g) = \sqrt{\frac1{N}\sum_{i=1}^N g_i^2}
$$

in the benchmark paper [2009.11662].

This protocol creates exact supervised pairs: the model input is contaminated EEG \(y\), and the training target is the clean EEG \(x\) [2009.11662]. For ocular artifact experiments in the original benchmark, **3400 EEG segments** and **3400 ocular artifact segments** are split into **80% training**, **10% validation**, and **10% test**, then expanded by random linear mixing across **10 SNR levels** from **\(-7\) dB to \(2\) dB** [2009.11662]. For myogenic artifact experiments, the EEG pool is upsampled to **512 Hz** to match EMG, and some EEG segments are randomly reused because there are fewer clean EEG than EMG segments [2009.11662].

A second normalization stage is applied at model-training time: both clean and contaminated signals are normalized by the standard deviation of the contaminated signal, in the intended form

$$
\hat{x} = \frac{x}{\sigma_y}, \qquad \hat{y} = \frac{y}{\sigma_y}
$$

with the standard deviation retained for rescaling the denoised output back to the original amplitude domain [2009.11662]. This contaminated-signal-based normalization was reused in later EEGDenoiseNet studies, including the EMG-focused “Novel CNN” work and the RetNet-based EEGDiR model [2010.11709], [2404.15289].

A central methodological consequence is that EEGDenoiseNet evaluates denoisers under a controlled additive model with known ground truth. This makes the benchmark well suited to supervised reconstruction studies, but also constrains conclusions to the semi-synthetic setting. Later work repeatedly notes that success on EEGDenoiseNet does not by itself establish generalization to real contaminated EEG, where artifact coupling may be nonlinear, electrode-dependent, or mixed across multiple artifact classes [2209.11980], [2411.13288], [2606.08594].

## 3. Original benchmark models and evaluation protocol

The original EEGDenoiseNet paper provides not only the dataset but also a benchmark protocol with four deep-learning model families: a **fully connected neural network (FCNN)**, a **simple CNN**, a **complex CNN**, and an **LSTM-based RNN** [2009.11662]. All are trained end-to-end with mean squared error under the mapping

$$
\tilde{x} = f(\hat{y}, \boldsymbol{\theta})
$$

from normalized contaminated input to normalized clean target [2009.11662]. Training used **Adam** with \(\alpha = 5e^{-5}\), \(\beta_1 = 0.5\), and \(\beta_2 = 0.9\), implemented in **Python 3.7** and **TensorFlow 2.2** on **two NVIDIA Tesla V100 GPUs** [2009.11662].

The FCNN has **4 hidden layers** with **ReLU** and dropout, with layer width equal to input length: **512** neurons for ocular denoising and **1024** neurons for myogenic denoising [2009.11662]. The simple CNN contains **4 one-dimensional convolution layers** with kernel **\(1 \times 3\)**, stride **1**, **64** feature maps, batch normalization, ReLU, and a flatten-plus-dense reconstruction head [2009.11662]. The complex CNN is a **1D-ResCNN** with parallel residual branches using kernels **\(1 \times 3\)**, **\(1 \times 5\)**, and **\(1 \times 7\)** [2009.11662]. The RNN benchmark is an **LSTM** whose outputs are passed through a three-layer fully connected network with ReLU and dropout [2009.11662].

Evaluation is based on convergence curves, temporal-domain relative RMSE, spectral-domain relative RMSE, and correlation coefficient. The original benchmark defines temporal and spectral error as

$$
RRMSE_{temporal} =\frac{RMS(f( y ) -x)}{RMS(x)}
$$

and

$$
RRMSE_{spectral} =\frac{RMS(PSD(f(y))-PSD(x))}{RMS(PSD(x))}
$$

with correlation measured by the covariance-normalized coefficient \(CC\) [2009.11662]. The benchmark also includes two traditional single-channel baselines, **empirical mode decomposition** and filtering, and reports repeated randomized experiments across **10** independently generated datasets for statistical robustness [2009.11662].

The benchmark’s headline findings are artifact-dependent. For ocular artifact removal, the **RNN** is reported to have the lowest RRMSE and the highest CC among the four deep models; for myogenic artifact removal, the **complex CNN** performs best overall [2009.11662]. The paper further reports that deep-learning methods outperform EMD and filtering for both ocular and myogenic denoising, with **\(p < 0.001\)** in ANOVA-based comparisons for temporal RRMSE, spectral RRMSE, and CC [2009.11662].

## 4. Benchmark behavior, artifact asymmetry, and methodological scope

A persistent feature of EEGDenoiseNet is the asymmetry between **ocular** and **myogenic** denoising. The original benchmark paper explicitly associates ocular artifacts with lower-frequency contamination and myogenic artifacts with broader, higher-frequency spectral structure [2009.11662]. Later studies sharpen this interpretation. The frequency-based deep-learning model of 2023 characterizes ocular activity as low-frequency, high-amplitude activity around **0.5–3 Hz**, whereas muscle activity overlaps more broadly with EEG and is usually **\(\le 35\) Hz**, making EMG denoising harder because it risks removing genuine neural content [2310.17335]. This suggests that EEGDenoiseNet does not represent a single homogeneous denoising task, but rather at least two benchmark regimes with different spectral separability and error profiles.

This asymmetry is visible across later models. The frequency-based dynamic-filtering approach reported in 2023 achieves its strongest result on ocular artifact removal, where it reports **\(RRMSE_t=0.405\)**, **\(RRMSE_f=0.490\)**, and **\(CC=0.917\)**, while remaining only competitive on muscle artifact removal with **\(RRMSE_t=0.573\)**, **\(RRMSE_f=0.496\)**, and **\(CC=0.805\)** [2310.17335]. EEGDiR likewise reports larger margins over prior methods on EOG and SS2016 EOG than on EMG, with average EEGDenoiseNet EOG results of **\(0.3279, 0.3616, 0.9329\)** for \(RRMSE_t\), \(RRMSE_s\), and \(CC\), versus **\(0.5322, 0.5004, 0.8072\)** on EMG [2404.15289].

The benchmark’s scope is also limited by construction. EEGDenoiseNet is built around **single-artifact contamination**, conventionally evaluated separately for EOG and EMG. Even when later work introduces “multi-artifact” denoisers, the term often refers to one shared model trained across separate EOG-corrupted and EMG-corrupted samples, not simultaneous contamination by both artifacts in the same input [2310.17335]. Only more recent work such as BandRouteNet explicitly constructs a **mixed EOG/EMG** dataset on top of EEGDenoiseNet and evaluates that case under a unified protocol [2604.24428]. The original benchmark therefore standardizes single-channel, single-artifact supervised denoising more than full real-world artifact ecology.

## 5. Subsequent model families built on EEGDenoiseNet

EEGDenoiseNet quickly became the standard evaluation setting for a broad range of neural denoisers. One important line of work sought stronger CNN baselines for the harder EMG task. The 2020 “Novel CNN” paper used **4514 EEG** and **5598 EMG** epochs from EEGDenoiseNet, expanded EEG by reuse to **5598 pairs**, split them into **4478/560/560** train/validation/test pairs, and trained a **7-block 1D CNN** with progressively increasing channels **32, 64, 128, 256, 512, 1024, 2048** and average pooling after the first six blocks [2010.11709]. On EEGDenoiseNet EMG denoising, it reported **\(RRMSE_t = 0.448\)**, **\(RRMSE_f = 0.442\)**, and **\(CC = 0.863\)**, outperforming the benchmark FCNN and RNN on average, while also arguing that the new architecture largely eliminated the severe overfitting observed in earlier CNN baselines [2010.11709].

A second line adapted sequence models. EEGDnet reshapes a **1-D EEG epoch** into a **\(k \times q\)** matrix, with baseline configuration **\(8 \times 64\)** for 512-sample ocular experiments, and applies a Transformer encoder to jointly model what the paper terms non-local and local self-similarity [2109.04235]. Under the EEGDenoiseNet protocol with **4514 EEG**, **3400 ocular**, and **5598 muscle** epochs, contaminated by \(y=x+\lambda n\) at uniformly distributed SNR from **\(-7\) dB to \(2\) dB**, EEGDnet reports for ocular artifact removal **\(\mathrm{RRMSE}_{temporal}=0.497\)**, **\(\mathrm{RRMSE}_{spectral}=0.491\)**, and **\(CC=0.868\)**, and for muscle artifacts **\(0.677\)**, **\(0.626\)**, and **\(0.732\)** respectively [2109.04235].

EEGDiR then replaced Transformer attention with a Retentive Network-derived mechanism, using **signal embedding** that patchifies a 512-sample segment with patch size **16**, yielding a **\(32 \times 512\)** hidden representation and **4 DiR blocks** with **8 heads** [2404.15289]. It retains the canonical contamination model \(y=x+\lambda n\) and EEGDenoiseNet’s **\(-7\) dB to \(2\) dB** range, and reports improved average scores over SCNN, 1D-ResCNN, LSTM, and EEGDnet on EOG, EMG, and SS2016 EOG [2404.15289].

Another line explored explicit frequency conditioning. The 2023 paper “A multi-artifact EEG denoising by frequency-based deep learning” treats EEGDenoiseNet as a controlled setting where the **PSD of the pure noise \(\lambda n\)** and the **PSD of the noisy signal \(y\)** can be used as side information to generate sample-specific denoising kernels [2310.17335]. The network’s kernel evaluator contains two symmetric branches of **1D convolutional layers with tanh activations** for the real and imaginary parts of a filter in the frequency domain, followed by **IRFFT**, a **kernel-size-1 1D convolution**, and final **tanh**, with total trainable parameter count **224,192** [2310.17335]. Its distinctiveness is not merely performance but the way it exploits EEGDenoiseNet’s synthetic construction to assume access to contaminant PSD.

Generative models have also been evaluated on EEGDenoiseNet. A 2024 study recasts single-channel EMG denoising as paired image-to-image translation with **pix2pixGAN**, converting each 1024-sample contaminated waveform into a **32×32 image**, training a conditional GAN, and reshaping the denoised image back to a time series [2411.13288]. Its average results over all SNR levels show **CC 0.8645**, **RRMSE\(_{\text{spectral}}\) 0.3596**, and **RRMSE\(_{\text{temporal}}\) 0.5797**, indicating particularly strong spectral recovery and correlation, though not the best average temporal RRMSE by its own table [2411.13288]. A more compact efficiency-oriented line is represented by TADA, a **Targeted Adversarial Denoising Autoencoder** with **less than 400,000 trainable parameters**, SNR-aware targeting, and covariance-guided logistic rescaling, evaluated on EEGDenoiseNet EMG removal from **67 subjects** [2501.04967].

The benchmark has also supported extensions beyond the original dataset. One ocular-artifact study combined EEGDenoiseNet with DEAP, yielding **6164 clean EEG segments**, **5000 EOG segments**, and a derived **150,000 augmented** sample corpus, and used **BiLSTM** models with raw, STFT, CWT, and WSST inputs, finding **WSST-Net** best with average MSE **0.3066** [2209.11980]. Although this differs from pure EEGDenoiseNet benchmarking, it shows the benchmark’s role as a base resource for constructing larger semi-synthetic denoising datasets.

## 6. Benchmark saturation, reproducibility issues, and current interpretation

Later analyses have shifted the interpretation of EEGDenoiseNet from merely a shared benchmark to an object of methodological scrutiny. The 2026 capacity-controlled study argues that the benchmark is already close to **saturated** with very small models [2606.08594]. Holding architecture family, loss, split, optimizer, and training recipe fixed, and varying only channel width in a depthwise-separable convolutional U-Net from **1.05K** to **40.26K** parameters, the study reports that reconstruction performance on EEGDenoiseNet saturates by roughly **3–6.5K parameters**, with post-elbow gains of at most **0.015 correlation coefficient per \(\log_{10}\)-parameter unit** [2606.08594]. A retrained **8.46M-parameter** EEGDN CNN matches the **40.26K** compact model on EOG, a roughly **200×** parameter gap without reconstruction advantage under the same controlled pipeline [2606.08594].

The same paper argues that strong EEGDenoiseNet reconstruction metrics do not reliably predict downstream BCI utility. Using denoisers trained under EEGDenoiseNet-style supervision and testing them on BCI Competition IV-2a motor imagery classification, it reports that reconstruction-optimized denoising can significantly degrade **CSP+LDA** accuracy across all nine subjects and multiple artifact conditions, including naturally recorded trials [2606.08594]. This does not invalidate EEGDenoiseNet as a reconstruction benchmark; rather, it narrows what the benchmark can be said to measure. A plausible implication is that EEGDenoiseNet is valuable for standardized supervised signal restoration, but insufficient as a sole proxy for preservation of task-relevant neural information.

Reproducibility is another persistent issue in EEGDenoiseNet-based literature. Many later papers note missing details in architectural specifications, data split procedures, exact PSD implementations, or subject-wise partitioning [2310.17335], [2411.13288], [2501.04967], [2604.24428]. In some studies, printed equations are malformed, especially normalization formulas, loss expressions, and correlation metrics [2310.17335], [2209.11980], [2404.15289]. BandRouteNet, despite strong reported results under a unified protocol and only **0.20M** parameters, leaves unspecified several implementation details such as batch size, learning-rate schedule, Inception1D branch structure, GRU dimensions, and attention head counts [2604.24428]. These issues have made “same-pipeline” retraining studies especially important for interpreting benchmark progress.

Taken together, EEGDenoiseNet is best understood as a standardized but constrained benchmark for **single-channel, semi-synthetic, supervised EEG denoising**, built around clean-reference reconstruction under controlled EOG and EMG contamination [2009.11662]. Its enduring significance lies in providing a common substrate on which denoising architectures can be compared, training regimes can be stress-tested, and broader questions about capacity, artifact-specific difficulty, and metric validity can be studied [2009.11662], [2606.08594]. Its limitations—single-artifact bias, additive mixing assumptions, lack of real noisy-clean pairs, and incomplete alignment with downstream utility—have become part of the benchmark’s identity rather than merely external caveats [2209.11980], [2411.13288], [2606.08594].

Source: https://www.emergentmind.com/topics/eegdenoisenet