Research on EEG Recognition Based on Deep Neural Networks with Attention Mechanism

DOI: https://doi.org/10.62517/jike.202304305

Author(s)

Yufeng Wang, Geng Fang, Binbin Chen, Zhengdong Chen*

Affiliation(s)

School of Information Engineering, Nanjing Polytechnic Institute, Nanjing, Jiangsu, China *Corresponding Author.

Abstract

Electroencephalogram(EEG) is widely applied in brain-computer interface systems. Due to its high time resolution, portability, and non-invasive characteristics, it has been applied for the rehabilitation of disabled patients. At present, deep learning has also been applied in EEG feature extraction and pattern classification. However, most EEG features in deep neural networks lack discriminative ability, which limits the representation ability of these methods. To overcome this issue, two methods were proposed, named ConvNet_S_A and ConvNet_S_CA, which are based on the backbone network Shallow ConvNet. The proposed methods combine the channel attention mechanism (SE module) and the convolutional block attention module (CBAM module) to learn discriminative EEG features, respectively. Experimental results on the public EEG datasets and self-collected EEG datasets demonstrate the effectiveness.

Keywords

Electroencephalogram; Motor Imagery; Deep Neural Networks; Attention Mechanism

References

[1]Lebedev M A, Nicolelis M. Brain-machine interfaces: past, present and future. Trends in Neurosciences, 2006, 29 (9): 536-546. [2]Hatem A, Peter B, Hotz B S, et al. What disconnection tells about motor imagery: evidence from paraplegic patients. Cerebral Cortex, 2004, 15 (2): 131. [3]Lecun Y, Bengio Y, Hinton G. Deep learning. Nature, 2015, 521 (7553): 436. [4]Huang J, Xu X, Zhang T. Emotion classification using deep neural networks and emotional patches//2017 IEEE International Conference on Bioinformatics and Biomedicine (BIBM). IEEE, 2017.DOI:10.1109/BIBM.2017.8217786. [5]Schirrmeiste R T, Springenberg J T, Fiedere L D J, et al. Deep learning with convolutional neural networks for EEG decoding and visualization. Human Brain Mapping, 2017. DOI:10.1002/hbm.23730. [6]Dornhege G. Combined optimization of spatial and temporal filters for improving brain-computer interfacing. IEEE transactions on bio-medical engineering, 2006, 53 (11): 2274. [7]Hinterberger T, A Kübler, Kaiser J, et al. A brain–computer interface (BCI) for the locked-in: comparison of different EEG classifications for the thought translation device. Clinical Neurophysiology, 2003, 114 (3): 416-425. [8]An X, Kuang D, Guo X, et al. A Deep Learning Method for Classification of EEG Data Based on Motor Imagery. Springer International Publishing, 2014. DOI:10.1007/978-3-319-09330-7_25. [9]Tabar Y R, Halici U. A novel deep learning approach for classification of EEG motor imagery signals. Journal of Neural Engineering, 2017, 14 (1): 016003. [10]Naeem M, Brunner C, Leeb R, et al. Seperability of four-class motor imagery data using independent components analysis. Journal of Neural Engineering, 2006, 3 (3): 208-216. [11]Hu J, Shen L, Albanie S, et al. Squeeze-and-Excitation Networks. IEEE Transactions on pattern analysis and machine intelligence, 2020, 42 (8) 2011-2023.DOI:10.1109/TPAMI. 2019. 2913372. [12]Woo S, Park J, Lee J Y, et al. CBAM: Convolutional Block Attention Module. Springer, Cham, 2018. DOI:10.1007/978-3-030-01234-2_1. [13]Hang W, Feng W, Liang S, et al. Deep stacked support matrix machine based representation learning for motor imagery EEG classification. Computer Methods and Programs in Biomedicine, 2020, 193: 105466. [14]Maaten L J P V D, Hinton G E. Visualizing High-Dimensional Data using t-SNE. Journal of Machine Learning Research, 2008, 9: 2579-2605.