Meta-Learning for EEG Motor Imagery Classification

Brain-computer interface (BCI) for motor imagery is an advanced technology used in the field of medical rehabilitation. However, due to the poor accuracy of electroencephalogram feature classification, BCI systems often misrecognize user commands. Although many state-of-the-art feature selection methods aim to enhance classification accuracy, they usually overlook the interrelationships between individual features, indirectly impacting the accuracy of feature classification. To overcome this issue, we propose an adaptive feature learning model that employs a Riemannian geometric approach to generate a feature matrix from electroencephalogram signals, serving as the model's input. By integrating the enhanced adaptive L1 penalty and weighted fusion penalty into the sparse learning model, we select the most informative features from the matrix. Specifically, we measure the importance of features using mutual information and introduce an adaptive weight construction strategy to penalize regression coefficients corresponding to each variable adaptively. Moreover, the weighted fusion penalty balances weight differences among correlated variables, reducing the model's overreliance on specific variables and enhancing accuracy. The performance of the proposed method was validated on BCI Competition IV datasets IIa and IIb using the support vector machine. Experimental results demonstrate the effectiveness and superiority of the proposed model compared to the existing models.

Brain-computer interface (BCI) based on electroencephalogram (EEG) is a promising technology, allowing computers to estimate human intentions. Intention recognition tool such as motor imagery (MI) with high reliability is one of the major challenges in the BCI field. Recently, researchers have attempted to use transfer learning for various BCI datasets, but the studies showed low classification accuracy. This study aimed to increase the classification accuracy of the MI through sequential transfer learning for a single dataset. EEG-MI data with 9 subjects from the dataset 2a of BCI competition IV were used. EEGNet was used for MI classification. The pre-trained model was constructed by first learning whether the data were MI or not. The model was then sequentially fine-tuned through transfer learning for four MI tasks (i.e., left hand, right hand, both feet and tongue). The model was able to classify MI with 91.34% accuracy. In the meantime, the baseline model without transfer learning showed an accuracy of 61.62%, whereas the fine-tuned model presented an improved accuracy of 63.82%. Consequently, the sequential transfer learning was able to improve the performance of MI-BCI.

Brain-Computer Interface (BCI) systems allow user to operate devices without muscular activation and experimental results indicate that they can induce activity-dependent plasticity. A BCI system has two key features: exploitation of brain signal changes induced by the cognitive task assigned to the user and mutual learning. With respect to the first feature, BCI users can learn to control and hence change their cortical activity, for example by performing a motor imagery task. Mutual learning, on the other hand, is set up between users, which have to learn how to control their brain activity, and the classifier used for learning and categorize brain signal changes. This mutual learning can be mediated by the use of feedback, which in turn can be discrete or continuous, abstract or realistic and so on. Motor imagery represents a promising approach for stroke rehabilitation, thus a BCI based on a motor imagery paradigm could be a useful tool. Nevertheless, in stroke rehabilitation settings two issues are emerging which are also crucial for motor imagery BCIs. The first one relates to the importance of recognize the imagery ability of the subject, the second one to the importance of feedback in re-learning of motor skills. The present study relates to this second issue and it is aimed at investigating attitudes of BCI users towards feedback. We considered a sample of six healthy males, using an EEG-based BCI and executing a Motor Imagery task without feedback (training phase) and with feedback (performance phase). A horizontal bar of varying length depicted on a computer screen was used as feedback medium and overall accuracy data were matched with neurophysiological data acquired from P z and O z . Results could be of importance for BCI use as neurorehabilitation tool in stroke rehabilitation protocols since they suggest that some subjects could need an increased cortical activation in order to deal with a stimulus like visual feedback. Moreover, they reveal a possible association between worsening in performance and increased cortical activation needed in order to cope with feedback.

Highly sophisticated control based on a brain-computer interface (BCI) requires decoding kinematic information from brain signals. The forearm is a region of the upper limb that is often used in everyday life, but intuitive movements within the same limb have rarely been investigated in previous BCI studies. In this study, we focused on various forearm movement decoding from electroencephalography (EEG) signals using a small number of samples. Ten healthy participants took part in an experiment and performed motor execution (ME) and motor imagery (MI) of the intuitive movement tasks (Dataset I). We propose a convolutional neural network using a channel-wise variational autoencoder (CVNet) based on inter-task transfer learning. We approached that training the reconstructed ME-EEG signals together will also achieve more sufficient classification performance with only a small amount of MI-EEG signals. The proposed CVNet was validated on our own Dataset I and a public dataset, BNCI Horizon 2020 (Dataset II). The classification accuracies of various movements are confirmed to be 0.83 (±0.04) and 0.69 (±0.04) for Dataset I and II, respectively. The results show that the proposed method exhibits performance increases of approximately 0.09~0.27 and 0.08~0.24 compared with the conventional models for Dataset I and II, respectively. The outcomes suggest that the training model for decoding imagined movements can be performed using data from ME and a small number of data samples from MI. Hence, it is presented the feasibility of BCI learning strategies that can sufficiently learn deep learning with a few amount of calibration dataset and time only, with stable performance.

Few-shot learning (FSL) aims to recognize new categories given limited training samples. The core challenge is to avoid overfitting to the minimal data while ensuring good generalization to novel classes. One mainstream method employs prototypes from visual feature extractors as classifier weight and the performance depends on the quality of the prototype. Since different categories may have similar visual features, the visual prototype has limitations. This is because existing methods only learn a simple visual feature extractor during the pre-training stage but neglect the importance of a well-developed feature space for the prototype. We introduce the Semantic Enhanced Visual Prototype framework (SEVpro) to address this issue. SEVpro refines prototype learning from the pre-training stage and serves as a versatile plug-and-play framework for all prototype-based FSL methods. Specifically, we enhance prototype discriminability by transforming semantic embeddings into the visual space, aiding in separating categories with similar visual features. For novel class learning, we leverage knowledge from base classes and incorporate semantic information to elevate prototype quality further. Meanwhile, extensive experiments on FSL benchmarks and ablation studies demonstrate the superiority of our proposed SEVpro for FSL.

EEG-based brain-computer interfaces (BCI) are non-invasive approaches for replacing or restoring motor functions in impaired patients, and direct brain-to-device communication in the general population. Motor imagery (MI) is one of the most used BCI paradigms, but its performance varies across individuals and certain users require substantial training to develop control. In this study, we propose to integrate a MI paradigm simultaneously with a recently proposed Overt Spatial Attention (OSA) paradigm, to accomplish BCI control. We evaluated a cohort of 25 human subjects' ability to control a virtual cursor in one- and two-dimensions over 5 BCI sessions. The subjects used 5 different BCI paradigms: MI alone, OSA alone, MI, and OSA simultaneously towards the same target (MI+OSA), and MI for one axis while OSA controls the other (MI/OSA and OSA/MI). Our results show that MI+OSA reached the highest average online performance in 2D tasks at 49% Percent Valid Correct (PVC), and statistically outperforms both MI alone (42%) and OSA alone (45%). MI+OSA had a similar performance to each subject's best individual method between MI alone and OSA alone (50%) and 9 subjects reached their highest average BCI performance using MI+OSA. Integrating MI and OSA leads to improved performance over both individual methods at the group level and is the best BCI paradigm option for some subjects. This work proposes a new BCI control paradigm that integrates two existing paradigms and demonstrates its value by showing that it can improve users' BCI performance.

Decoding brain activity for brain computer interfaces (BCI) is a challenging machine learning task. Many BCIs are designed in a synchronous setting, which allows to output a command from the brain to the computer at pre-determined, specific time points. The task becomes substantially more demanding when designing asynchronous BCIs, which allow control over the interface at any given time. Most BCIs use static classifiers, in which the readout is driven only by the current input. However, due to the complex nature of ongoing brain activity, dynamic classifiers, which capture the temporal dynamics of the input signal, are likely to be more suitable for the task. To examine this notion, we designed a classification framework based on Echo State Networks (ESNs), which have demonstrated strong performance in similar tasks in other domains. We evaluated the performance of this approach on pre-recorded EEG data of an asynchronous motor imagery (MI) BCI task, and compared our approach to a conventional method. ESN-based architectures exhibited superior performance in comparison. Our findings suggest that ESNs may prove useful in other BCI paradigms in virtue of their ability to reduce the detrimental effect of non-stationarity in brain signals.

The development of brain computer interfaces (BCIs) for patient rehabilitation is a growing field of research. The BCI experimental paradigms consist mainly of selective attention BCI models and motor imagery (MI) BCIs. Selective attention models require an external stimulus (screen) but achieve high rates of classification accuracy fairly quickly. MI systems do not require external stimuli but require extensive training. The goal of our study was to gauge how much a short training paradigm requiring seven six-minute sessions of video attention would change fMRI BOLD activity between two sessions of MI ± one acquired before and another after the training protocol took place. The study used four MIs: 1) right hand 2) left hand 3) feet 4) tongue; and it was carried out on ten subjects. We found an increase of the BOLD response after training, both in amplitude and spatial extent, for the majority (6 out of 10) of subjects and MIs (all except left hand). Our results corroborate other literature results regarding the effect of training in MI based BCIs.

In the aging society, the number of people suffering from vascular disorders is rapidly increasing and has become a social problem. The death rate due to stroke, which is the second leading cause of global mortality, has increased by 40% in the last two decades. Stroke can also cause paralysis. Of late, brain‐computer interfaces (BCIs) have been garnering attention in the rehabilitation field as assistive technology. A BCI for the motor rehabilitation of patients with paralysis promotes neural plasticity, when subjects perform motor imagery (MI). Feedback, such as visual and proprioceptive, influences brain rhythm modulation to contribute to MI learning and motor function restoration. Also, virtual reality (VR) can provide powerful graphical options to enhance feedback visualization. This work aimed to improve immersive VR‐BCI based on hand MI, using visual‐electrotactile stimulation feedback instead of visual feedback. The MI tasks include grasping, flexion/extension, and their random combination. Moreover, the subjects answered a system perception questionnaire after the experiments. The proposed system was evaluated with twenty able‐bodied subjects. Visual‐electrotactile feedback improved the mean classification accuracy for the grasping (93.00% ± 3.50%) and flexion/extension (95.00% ± 5.27%) MI tasks. Additionally, the subjects achieved an acceptable mean classification accuracy (maximum of 86.5% ± 5.80%) for the random MI task, which required more concentration. The proprioceptive feedback maintained lower mean power spectral density in all channels and higher attention levels than those of visual feedback during the test trials for the grasping and flexion/extension MI tasks. Also, this feedback generated greater relative power in the μ‐band for the premotor cortex, which indicated better MI preparation. Thus, electrotactile stimulation along with visual feedback enhanced the immersive VR‐BCI classification accuracy by 5.5% and 4.5% for the grasping and flexion/extension MI tasks, respectively, retained the subject’s attention, and eased MI better than visual feedback alone.

Brain Computer Interface (BCI) has become one of the most interesting alternatives to support automatic systems able to interpret brain functions. Recently, the Motor Imagery (MI) paradigm is a widely topic of interest as a tool to develop BCI-based systems. Here, we present a relevant feature extraction methodology, termed MI discrimination using kernel relevance analysis (MIDKRA), to support MI classification in BCI systems. For such a purpose, a similarity criterion to rank the contribution of EEG features for classifying an MI paradigm is employed. The introduced approach includes the supervised information regarding the MI paradigm to find out a relevant set of features encoding discriminative information. We model the EEG recordings by considering both time and time-frequency representations. Moreover, a k nearest-neighbor classifier is carried out to validate the proposed feature relevance approach. Experimental results carried out on two different BCI databases, a well-known public MI data and an Emotiv-based dataset built by us, demonstrate that MIDKRA outperforms state of the art methods and it is a suitable alternative to support straightforward BCI systems.

With the continuous development of the brain-computer interface (BCI) technology, the lower limb rehabilitation system based on Motor Imagery (MI) has gradually become a research hotspot in the field of rehabilitation. To recognize the lower limbs MI, this paper designed an experimental paradigm for MI lower limb MI and used the 1D-CNN-LSTM deep learning algorithm to classify lower limb movement features from MI EEG signals. Compared with classical machine learning algorithms, the results showed that 1D-CNN-LSTM has relatively higher accuracy. Meanwhile, the paper built a real-time lower limb rehabilitation system based on the 1D-CNN-LSTM algorithm, which verifies the effectiveness and feasibility of the algorithm. The system provides an advanced and effective solution for brain-computer interfaces based on MI.

MetAdapt: Meta-learned task-adaptive architecture for few-shot classification

For individuals with high degrees of motor disability or locked-in syndrome, it is impractical or impossible to use mechanical switches to interact with electronic devices. Brain computer interfaces (BCIs) can use motor imagery to detect interaction intention from users but lack the accuracy of mechanical switches. Hence, there exists a strong need to improve the accuracy of EEG-based motor imagery BCIs attempting to implement an on/off switch. Here, we investigate how monitoring the pupil diameter of a person as a psycho-physiological parameter in addition to traditional EEG channels can improve the classification accuracy of a switch-like BCI. We have recently noticed in our lab (work not yet published) how motor imagery is associated with increases in pupil diameter when compared to a control rest condition. The pupil diameter parameter is easily accessible through video oculography since most gaze tracking systems report pupil diameter invariant to head position. We performed a user study with 30 participants using a typical EEG based motor imagery BCI. We used common spatial patterns to separate motor imagery, signaling movement intention, from a rest control condition. By monitoring the pupil diameter of the user and using this parameter as an additional feature, we show that the performance of the classifier trying to discriminate motor imagery from a control condition improves over the traditional approach using just EEG derived features. Given the limitations of EEG to construct highly robust and reliable BCIs, we postulate that multi-modal approaches, such as the one presented here that monitor several psycho-physiological parameters, can be a successful strategy in making BCIs more accurate and less vulnerable to constraints such as requirements for long training sessions or high signal to noise ratio of electrode channels.

This study establishes the effectiveness of event related synchronisation (ERS) features for a system paced brain computer interface (BCI). In particular, the relationship between the duration of motor imagery (MI) and the quality of the features extracted from the ERS is investigated. To this end, two groups of users performed brief (2s) or sustained (4s) MI, and offline single trial BCIs were validated on each group based on features extracted from the EEG before, during and after MI. The BCIs were designed to recognise two intentional control tasks and a no-control state. Cross-validated results indicate that brief MI leads to more informative ERS features than sustained MI.

ABSTRACT
\n
\nPutri, Anita Kurnia. 2020. “Analisis Kontrastif Adverbia Tinkatan Penanda Negasi dalam Bahasa Jepang dan Bahasa Jepang”. Thesis, department of Japanese Language and Culture, Faculty of Humanities, Diponegoro University. The Advisor: Lina Rosliana, S.S., M.Hum.
\n
\nThe problems that will be discussed in this thesis include: 1. What is the structure and meaning of the adverb in the level of negation markers in Japanese sentences? 2. What is the structure and meaning of the adverb at the level of the negation marker in Indonesian sentences? 3. How are the adverbs of the level of negation markers similar and different in Japanese and Indonesian sentences?
\n
\nMeanwhile, the purpose of this study is to describe the structure and meaning of adverbs in Japanese and Indonesian sentences. After knowing the structure and meaning, we will examine the similarities and differences in the structure and meaning of the adverbial levels of the negation markers of the two languages.
\n
\nIn analyzing the data, this study used the agih method followed by contrastive techniques. The data were obtained from the Japanese language website and news articles in Indonesian. Then the data is presented in words or descriptive.
\n
\nBased on the research, the structural similarities are both followed by verbs, adjectives. While the difference, adverbs in Japanese can be followed by phrases or clauses and negative forms, Indonesian adverbial followed by nouns. In meaning, the adverbs of the level of negation markers in Japanese and Indonesian indicates the meaning of the situation, the action is under adequate. And in terms of meaning it has 7 differences.
\n
\nKeywords : contrastive, adverbs, teido no fukushi
\n