An end-to-end deep learning approach to MI-EEG signal classification for BCIs

Dose, Hauke; Møller, Jakob Skadkær; Iversen, Helle K.; Puthusserypady, Sadasivan

doi:10.1016/j.eswa.2018.08.031

Cited by 269 publications

(214 citation statements)

References 39 publications

(59 reference statements)

Supporting

Mentioning

208

Contrasting

Unclassified

Order By: Relevance

“…Previous methods exploit shallow networks to match the domains of a single level; deep neural networks are good at extracting multilevel and compact features and will have better descriptions for specific tasks [26][27][28][29][30][31][32][33][34]. Many deep learning approaches are applied to decode EEG signals.…”

Section: Related Workmentioning

confidence: 99%

“…Deep learning has been proved not only to have more power to extract compact and deep-level features but also possess more strength to represent the task. It has won great achievements in many fields, especially including EEG decoding [21][22][23][24][25][26][27][28][29][30][31][32][33][34]. For instance, we focus anomaly detection [21], visual evoked potentials [22], P300 detection [24], workload analysis [25], error-related negativity responses (ERN) [30], movement-related cortical potentials (MRCP) [30], attentional information [35], and motor imagery tasks [26][27][28][29][30][31][32][33][34].…”

Section: Introductionmentioning

confidence: 99%

“…Usually, data from different subjects are pooled together and fed directly into the network, regardless of the statistical distribution discrepancy across subjects. It will result in obvious deterioration of the network [7,34,36].…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Deep Neural Network with Joint Distribution Matching for Cross-Subject Motor Imagery Brain-Computer Interfaces

Zhao

Liu

et al. 2020

BioMed Research International

View full text Add to dashboard Cite

Motor imagery brain-computer interfaces (BCIs) have demonstrated great potential and attract world-spread attentions. Due to the nonstationary character of the motor imagery signals, costly and boring calibration sessions must be proceeded before use. This prevents them from going into our realistic life. In this paper, the source subject's data are explored to perform calibration for target subjects. Model trained on source subjects is transferred to work for target subjects, in which the critical problem to handle is the distribution shift. It is found that the performance of classification would be bad when only the marginal distributions of source and target are made closer, since the discriminative directions of the source and target domains may still be much different. In order to solve the problem, our idea comes that joint distribution adaptation is indispensable. It makes the classifier trained in the source domain perform well in the target domain. Specifically, a measure for joint distribution discrepancy (JDD) between the source and target is proposed. Experiments demonstrate that it can align source and target data according to the class they belong to. It has a direct relationship with classification accuracy and works well for transferring. Secondly, a deep neural network with joint distribution matching for zero-training motor imagery BCI is proposed. It explores both marginal and joint distribution adaptation to alleviate distribution discrepancy across subjects and obtain effective and generalized features in an aligned common space. Visualizations of intermediate layers illustrate how and why the network works well. Experiments on the two datasets prove the effectiveness and strength compared to outstanding counterparts.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Deep Neural Network with Joint Distribution Matching for Cross-Subject Motor Imagery Brain-Computer Interfaces

Zhao

Liu

et al. 2020

BioMed Research International

View full text Add to dashboard Cite

show abstract

“…Multi-Adapt and MKAL have had comparable performance at that time even though these models do not capture the available temporal information in the time-series data. (Dose et al, 2018) builds a BMI and investigates DA for multi-variate EEG time-series data classification. The time-series classification of the multivariate EEG signals is a very similar challenge to the multi-variate sEMG signals.…”

Section: Source Data-absentmentioning

confidence: 99%

“…The time-series classification of the multivariate EEG signals is a very similar challenge to the multi-variate sEMG signals. (Dose et al, 2018) captures both the spatial and temporal correlations in the data with a CNN architecture. However, the DA is about supervised fine-tuning of all the model parameters on the target subject (such as (Donahue et al, 2014)) which is suboptimal as highlighted by (Du et al, 2017;Ketyk et al, 2019).…”

Section: Source Data-absentmentioning

confidence: 99%

On the Metrics and Adaptation Methods for Domain Divergences of sEMG-based Gesture Recognition

Ketykó¹,

Kovács²

2020

Proceedings of the 13th International Joint Conference on Biomedical Engineering Systems and Technologies

View full text Add to dashboard Cite

We propose a new metric to measure domain divergence and a new domain adaptation method for time-series classification. The metric belongs to the class of probability distributions-based metrics, is transductive, and does not assume the presence of source data samples. The 2-stage method utilizes an improved autoregressive, RNN-based architecture with deep/non-linear transformation. We assess our metric and the performance of our model in the context of sEMG/EMG-based gesture recognition under inter-session and inter-subject domain shifts.

show abstract

Rehabilitation Based on BCI: An Innovative Enhancement for Sensorimotor Cortex Rhythms Systemization

M.,

2024

Advanced Biology

View full text Add to dashboard Cite

The research proposes a novel strategy for categorizing electroencephalograms (EEG) in real‐time brain‐computer interfaces that have rehabilitation applications. The methodology utilizes Five Cross‐Common Spatial Patterns (FCCSP) to develop a motor movement/imagery systemization model that extracts multi‐domain characteristics with excellent performance. The goal is to eliminate the impact caused by EEG's nonstationarity. The article highlights the findings of a real‐time technique that is incorporated into a comprehensive prediction system, and it offers an innovative method to boost accuracy in real‐time Sensory‐Motor cortex Rhythms (SMR). The accuracy increased from 57.14% using raw EEG to 85.71% after preprocessing, and from 58.08% to 97.94% in public domain SMR. The proposed Butterworth bandpass filter is optimized using the FCCSP to determine the ideal bandwidth that incorporates the whole EEG features in beta waves. The Hybrid Systemization of the Correlated Feature Removal classifier is then integrated with the FCCSP method to create improved predictive models. As a consequence, while applied to real‐time and PhysioNet datasets, the outcome system achieved outstanding accuracy values of 85.71% and 97.94%, respectively. This demonstrates the robustness of the strategy to increase SMR prediction efficiency.

show abstract

An end-to-end deep learning approach to MI-EEG signal classification for BCIs

Cited by 269 publications

References 39 publications

Deep Neural Network with Joint Distribution Matching for Cross-Subject Motor Imagery Brain-Computer Interfaces

Deep Neural Network with Joint Distribution Matching for Cross-Subject Motor Imagery Brain-Computer Interfaces

On the Metrics and Adaptation Methods for Domain Divergences of sEMG-based Gesture Recognition

Rehabilitation Based on BCI: An Innovative Enhancement for Sensorimotor Cortex Rhythms Systemization

Contact Info

Product

Resources

About