Deep learning networks have been increasingly attracting attention in many fields. Recently,the application of deep learning models has been brought to the field of electroencephalographydenoising, and has provided performance that is comparable to that of traditional techniques.However, the lack of well-structured, standardized dataset with benchmark limits the develop-ment of deep learning solutions for EEG denoising. Therefore, we present EEGdenoiseNet, abenchmark dataset, that is suited for training and testing deep learning-based EEG denoisingmodels, as well as for comparing the performance across different models. Our EEGdenoiseNetdataset contains 4514 clean EEG epochs, 3400 EOG epochs and 5598 EMG epochs, whichallow users producing a large number of noisy EEG epochs with ground truth for modeltraining and testing. EEGdenoiseNet also offers a set of benchmarks generated by evaluatingthe performance of four classical deep learning networks (a fully-connected network, a simple convolution network, a complex convolution network and a recurrent neural network). Ourbenchmark dataset would hopefully accelerate the development of the emerging field of deeplearning-based EEG denoising .

For more information, The paper of this dataset is publicly available on arXiv(https://arxiv.org/abs/2009.11662).

Due to size limitations, EEG and EMG epochs with a sample rate of 512hz are temporarily placed in the G-node database (https://gin.g-node.org/NCClab/EEGdenoiseNet).

Single-Channel-EEG-Denoise tool box could be find in Github(https://github.com/ncclabsustech/Single-Channel-EEG-Denoise)

• [tensorflow] version = 2.2
• Python version = 3.6

Our laboratory also proposed an deep learning framework to separate neural signal and artifacts in the embedding space and reconstruct the denoised signal, which is called DeepSeparator. Could be find in Github(https://github.com/ncclabsustech/DeepSeparator).