Abstract:Electroencephalogram (EEG) has shown a useful approach to produce a brain-computer interface (BCI). One-dimensional (1-D) EEG signal is yet easily disturbed by certain artifacts (a.k.a. noise) due to the high temporal resolution. Thus, it is crucial to remove the noise in received EEG signal. Recently, deep learning-based EEG signal denoising approaches have achieved impressive performance compared with traditional ones. It is well known that the characteristics of self-similarity (including non-local and local ones) of data (e.g., natural images and time-domain signals) are widely leveraged for denoising. However, existing deep learning-based EEG signal denoising methods ignore either the non-local self-similarity (e.g., 1-D convolutional neural network) or local one (e.g., fully connected network and recurrent neural network). To address this issue, we propose a novel 1-D EEG signal denoising network with 2-D transformer, namely EEGDnet. Specifically, we comprehensively take into account the non-local and local self-similarity of EEG signal through the transformer module. By fusing non-local self-similarity in self-attention blocks and local self-similarity in feed forward blocks, the negative impact caused by noises and outliers can be reduced significantly. Extensive experiments show that, compared with other state-of-the-art models, EEGDnet achieves much better performance in terms of both quantitative and qualitative metrics.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: how to effectively remove noise in electroencephalogram (EEG) signals, especially ocular artifacts and muscle artifacts. Since one - dimensional (1 - D) EEG signals have high - time resolution and are easily affected by complex noise, effective denoising is crucial for improving the performance of EEG - based brain - computer interface (BCI) systems. Existing deep - learning methods, when dealing with EEG signal denoising, often consider only one of non - local self - similarity or local self - similarity, while ignoring the combination of the two. To overcome this limitation, this paper proposes a new 1 - D EEG signal denoising network based on 2 - D transformers - EEGDnet, which can simultaneously utilize non - local self - similarity and local self - similarity to reduce the impact of noise, thereby significantly improving the denoising effect. ### Main Contributions 1. **Fusion of Non - local and Local Self - similarity**: EEGDnet is the first deep - learning denoising method that comprehensively considers the non - local and local self - similarities of EEG signals. 2. **Introduction of 2 - D Transformers**: EEGDnet successfully applies 2 - D transformers to the denoising task of 1 - D time - domain signals, which is an innovative attempt. 3. **Low Resource Consumption**: EEGDnet has low model parameters and computational consumption on embedded BCI devices, and is suitable for practical applications. ### Method Overview The main framework of EEGDnet is shown in Figure 1. It is an end - to - end model. The input is a noisy EEG signal, and the output is a denoised EEG signal. The specific steps are as follows: 1. **Reshaping Layer**: Reshape the 1 - D EEG signal into a 2 - D matrix in order to utilize non - local self - similarity. 2. **Self - Attention Block**: Capture the non - local self - similarity of the signal through the self - attention mechanism. 3. **Feed - forward Block**: Process each segment through fully - connected layers and non - linear activation functions to utilize local self - similarity. 4. **Normalization Layer**: Apply layer normalization and residual connections after each block to reduce training time and improve model stability. 5. **Reshaping Layer**: Convert the 2 - D matrix back into a 1 - D signal as the final output. ### Experimental Results The experimental results show that EEGDnet outperforms other state - of - the - art methods in both quantitative and qualitative indicators. Especially in removing ocular artifacts and muscle artifacts, the performance of EEGDnet is significantly improved. In addition, EEGDnet also performs well in terms of computational amount, number of parameters, and storage size, and is suitable for deployment on resource - limited embedded devices. ### Conclusion The EEGDnet model proposed in this paper effectively solves the noise problem in EEG signals by fusing non - local and local self - similarities, especially in removing ocular artifacts and muscle artifacts. This model not only outperforms existing methods in performance, but also has an advantage in resource consumption, and is suitable for practical brain - computer interface systems.

EEGDnet: Fusing Non-Local and Local Self-Similarity for 1-D EEG Signal Denoising with 2-D Transformer

EEGDnet: Fusing non-local and local self-similarity for EEG signal denoising with transformer

EEGDiR: Electroencephalogram denoising network for temporal information storage and global modeling through Retentive Network

DTP-Net: Learning to Reconstruct EEG signals in Time-Frequency Domain by Multi-scale Feature Reuse

A GAN Guided Parallel CNN and Transformer Network for EEG Denoising.

A Novel Multimodule Neural Network for EEG Denoising

A general dual-pathway network for EEG denoising

Integrating spatial and temporal features for enhanced artifact removal in multi-channel EEG recordings

Exploiting Asymmetric EEG Signals with EFD in Deep Learning Domain for Robust BCI

EEGdenoiseNet: a benchmark dataset for deep learning solutions of EEG denoising

EEGdenoiseNet: A benchmark dataset for end-to-end deep learning solutions of EEG denoising

EEG Enhancement by Auto DNNs with Regularization of Spatial Feature Loss

A Dual-Branch Interactive Fusion Network to Remove Artifacts from Single-Channel EEG

EEGDfus: A Conditional Diffusion Model for Fine-Grained EEG Denoising

A Novel End-to-end Network Based on a bidirectional GRU and a Self-Attention Mechanism for Denoising of Electroencephalography Signals

Two-stage ECG Signal Denoising Based on Deep Convolutional Network.

A multi-artifact EEG denoising by frequency-based deep learning

An Efficient EEG Signal Fading Processing Framework Based on the Cognitive Limbic System and Deep Learning

Transformer-based Spatial-Temporal Feature Learning for EEG Decoding

EEG-Deformer: A Dense Convolutional Transformer for Brain-computer Interfaces