Improving brain–computer interface classification using adaptive common spatial patterns

Song, Xiaomu; Yoon, Suk-Chung

doi:10.1016/j.compbiomed.2015.03.023

Cited by 48 publications

(31 citation statements)

References 42 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Transfer learning methods are designed to make the training data domain closer to the test domain and have got great achievements in many areas, such as image, audio, and text processing [19,20,[37][38][39][40][41]. Due to the timevarying and nonstationary characters of EEG signals, the training data are statistically different from the test data, and the performance of the classifier obtained from the training data will degrade significantly, especially when test data come from different subjects [7][8][9][10][11][14][15][16][17][18]. Considering these, transfer learning methods are adopted to make improvements for BCI.…”

Section: Related Workmentioning

confidence: 99%

“…To overcome this limitation, several groups introduced machine learning, especially transfer learning methods for adapting BCIs to target subjects [7][8][9][10][11][14][15][16][17][18]. In recent years, several groups have started explicitly modelling such variations to exploit the common structure that is shared between multiple subjects.…”

Section: Introductionmentioning

confidence: 99%

“…In recent years, several groups have started explicitly modelling such variations to exploit the common structure that is shared between multiple subjects. Several works explored data from other subjects (called source subjects, the data from them are called source data), in order to regularize common spatial patterns (CSP), ultimately to make the estimation of covariance matrix more unbiased and filters more effective for target subjects (the data of them are called target data) [14][15][16]. Other works constructed filter bank to extract more abundant features, selected them according to some designed rules, and then ensembled them to obtain high performance [8,17,18].…”

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%

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

“…The system is a cue-paced BCI using MI of left versus right hand to control the flexion-extension of a 1 DOF-modelled arm on a screen, including a simple adaptive scheme based on CSP filtering and SVM classification [12]. The implemented adaptive scheme is similar to some previously proposed algorithms [21–24], generally defined as ACSP (adaptive common spatial pattern): we did not include these works among the previously cited ones [5, 7–16] because they have a different aim, that is, generically dealing with EEG inter- and intrasubject nonstationarities, rather than improving the user training process. However, beyond the implemented adaptive strategy, the system we describe was conceived as a whole, from training phase to utilization, and therefore includes a short calibration module without feedback (less than 3 minutes), followed by several repetitions of an adaptive module with feedback.…”

Section: Introductionmentioning

confidence: 99%

EEG-Based BCI System Using Adaptive Features Extraction and Classification Procedures

Mondini

Mangia

Cappello

2016

Computational Intelligence and Neuroscience

View full text Add to dashboard Cite

Motor imagery is a common control strategy in EEG-based brain-computer interfaces (BCIs). However, voluntary control of sensorimotor (SMR) rhythms by imagining a movement can be skilful and unintuitive and usually requires a varying amount of user training. To boost the training process, a whole class of BCI systems have been proposed, providing feedback as early as possible while continuously adapting the underlying classifier model. The present work describes a cue-paced, EEG-based BCI system using motor imagery that falls within the category of the previously mentioned ones. Specifically, our adaptive strategy includes a simple scheme based on a common spatial pattern (CSP) method and support vector machine (SVM) classification. The system's efficacy was proved by online testing on 10 healthy participants. In addition, we suggest some features we implemented to improve a system's “flexibility” and “customizability,” namely, (i) a flexible training session, (ii) an unbalancing in the training conditions, and (iii) the use of adaptive thresholds when giving feedback.

show abstract

“…Vol [15][16][17][18][19] The reduced system consists of 2 basic steps; EEG brain signal features extraction and their appropriate classification. 20,21 Assumed nature of the EEG signal on the one hand and how its impression from motor imaginary on the other hand, specifies what features of the EEG signal should be extracted and what an appropriate classifier should be used.…”

mentioning

confidence: 99%

An Effective Brain-Computer Interface System Based on the Optimal Timeframe Selection of Brain Signals

Abbaspour

Mehrshad

Razavi

2018

Int Clin Neurosci J

View full text Add to dashboard Cite

Introduction A brain-computer interface (BCI) system enables nerve and muscle free interaction of human with surrounding.1 Some recent and important research applications of BCI are human to human interface, neuroprosthetics, exoskeleton control, mobile and guided robotics, biometrics, neurogaming and Intelligent Transportation. [2][3][4][5] Due to the wide range of applications, BCI systems have the potential to use by both normal and disabled individual.6 Using this technology, severely disabled, paralyzed or who have neuromuscular diseases such as amyotrophic lateral sclerosis, stroke or spinal cord injuries become self-sufficient in fulfilling their basic requirements and severe motor disabilities. 7,8 Actually, BCI systems increase the quality of their life while reduce the burden and cost of care. Came the request in mind, creates a unique brain signal which is recognizable for an intelligent computational system. 10,11 Various techniques have been developed to extract brain signals which include magneto electroencephalography (MEG), functional magnetic resonance imaging (fMRI), near infrared spectroscopy (NIRS), electrocorticography (ECoG) and electroencephalography (EEG).2 EEG has advantages over other techniques, the most important of which is its good temporal resolution.12 Systems for recording EEG signal are also non-invasive, inexpensive, free of any radiation, and can be simply implemented.2 Thus, in order to capture motor imaginary brain activities in a BCI system, the EEG signal is commonly used.13 An EEG-based BCI system uses EEG as the control signal in neural and muscular free interaction of human with surrounding.1 In different parts of the EEG-based BCI system, the user's EEG signals captured by the electrodes and to decode and recognize the intended interactions are sent to the processor. 14The main purpose of EEG-based BCI is to translate EEG signal into a command for a computer.14 Generally, the design of such a system is very complex. 9 In many researches that have been done so far on EEG-based BCI systems, actually the system is reduced to a classifier

show abstract

Improving brain–computer interface classification using adaptive common spatial patterns

Cited by 48 publications

References 42 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

EEG-Based BCI System Using Adaptive Features Extraction and Classification Procedures

An Effective Brain-Computer Interface System Based on the Optimal Timeframe Selection of Brain Signals

Contact Info

Product

Resources

About