HS-CNN: a CNN with hybrid convolution scale for EEG motor imagery classification

Abstract

Objective. Electroencephalography (EEG) motor imagery classification has been widely used in healthcare applications such as mobile assistive robots and post-stroke rehabilitation. Recently, EEG motor imagery classification methods based on convolutional neural networks (CNNs) have been proposed and have achieved relatively high classification accuracy. However, these methods use single convolution scale in the CNN, while the best convolution scale differs from subject to subject. This limits the classification accuracy. Another issue is that the classification accuracy degrades when training data is limited. Approach. To address these issues, we have proposed a hybrid-scale CNN architecture with a data augmentation method for EEG motor imagery classification. Main results. Compared with several state-of-the-art methods, the proposed method achieves an average classification accuracy of 91.57% and 87.6% on two commonly used datasets, which outperforms several state-of-the-art EEG motor imagery classification methods. Significance. The proposed method effectively addresses the issues of existing CNN-based EEG motor imagery classification methods and improves the classification accuracy.

Data augmentation for deep-learning-based electroencephalography

“…They achieved 90% average accuracy, but did not report their multiplication factor or the accuracy before DA. [111]. They varied the CNN kernel size between subjects and even between sessions.…”

“…Figure 2A suggests that motor imagery was the most popular EEG task for DA in 2019. This led us to select the 2008 BCI competition IV dataset 2a as our motor imagery example [111,[117][118][119]. We selected left and right, as this was the most common classification technique in the literature and thus facilitated the comparison of our result with the literature [111].…”

“…Looking through the literature, the best accuracy we could find on the BCI competition IV dataset 2a dataset was for from a study by Dai and colleagues [111]. We therefore compared the accuracy for all 9 subjects in the dataset between that paper, our analysis without data augmentation, and our analysis with data augmentation ( Table 5).…”

“…Given the above, and to further test the usefulness of DA for EEG, we ran our own EEG analysis of an open dataset (BCI IV-2a), based on what we learned from our review of the literature. We then compared the accuracy were able to obtain with DA to the best accuracy we could find in the literature on the same dataset (Dai and colleagues [111]) as well as to our technique without DA. We were able to explain 24% of the accuracy beyond that explained by Dai et al, which suggests that our analysis of the literature is useful in gaining insight into which DA techniques work well for EEG analysis.…”

Cross-Dataset Variability Problem in EEG Decoding With Deep Learning

“…The lack of data problems makes deep learning models easily overfit. Some data augmentation methods may alleviate the overfitting problem for within-subject classification tasks (Wang et al, 2018;Dai et al, 2020). For cross-subject classification tasks, an easier way is to train the model directly on the entire dataset regardless of subject-specific information (Schirrmeister et al, 2017;Lawhern et al, 2018).…”