Improving the Performance of Individually Calibrated SSVEP-BCI by Task- Discriminant Component Analysis

IEEE Transactions on Neural Systems and Rehabilitation Engineering, Journal Year: 2021, Volume and Issue: 29, P. 1998 - 2007

Published: Jan. 1, 2021

A brain-computer interface (BCI) provides a direct communication channel between brain and an external device. Steady-state visual evoked potential based BCI (SSVEP-BCI) has received increasing attention due to its high information transfer rate, which is accomplished by individual calibration for frequency recognition. Task-related component analysis (TRCA) recent state-of-the-art method individually calibrated SSVEP-BCIs. However, in TRCA, the spatial filter learned from each stimulus may be redundant temporal not fully utilized. To address this issue, paper proposes novel method, i.e., task-discriminant (TDCA), further improve performance of individually-calibrated SSVEP-BCI. The TDCA was evaluated two publicly available benchmark datasets, results demonstrated that outperformed ensemble TRCA other competing methods significant margin. An offline online experiment testing 12 subjects validated effectiveness TDCA. present study new perspective designing decoding SSVEP-BCI presents insight implementation high-speed speller applications.

Language: Английский

---

SincNet-Based Hybrid Neural Network for Motor Imagery EEG Decoding

IEEE Transactions on Neural Systems and Rehabilitation Engineering, Journal Year: 2022, Volume and Issue: 30, P. 540 - 549

Published: Jan. 1, 2022

It is difficult to identify optimal cut-off frequencies for filters used with the common spatial pattern (CSP) method in motor imagery (MI)-based brain-computer interfaces (BCIs). Most current studies choose filter cut-frequencies based on experience or intuition, resulting sub-optimal use of MI-related spectral information electroencephalography (EEG). To improve utilization, we propose a SincNet-based hybrid neural network (SHNN) MI-based BCIs. First, raw EEG segmented into different time windows and mapped CSP feature space. Then, SincNets are as bank band-pass automatically data. Next, squeeze-and-excitation modules learn sparse representation filtered The data were fed convolutional networks deep representations. Finally, these features gated recurrent unit module seek sequential relations, fully connected layer was classification. We BCI competition IV datasets 2a 2b verify effectiveness our SHNN method. mean classification accuracies (kappa values) 0.7426 (0.6648) dataset 0.8349 (0.6697) 2b, respectively. statistical test results demonstrate that can significantly outperform other state-of-the-art methods datasets.

Language: Английский

---

Global Adaptive Transformer for Cross-Subject Enhanced EEG Classification

IEEE Transactions on Neural Systems and Rehabilitation Engineering, Journal Year: 2023, Volume and Issue: 31, P. 2767 - 2777

Published: Jan. 1, 2023

Due to the individual difference, EEG signals from other subjects (source) can hardly be used decode mental intentions of target subject. Although transfer learning methods have shown promising results, they still suffer poor feature representation or neglect long-range dependencies. In light these limitations, we propose Global Adaptive Transformer (GAT), an domain adaptation method utilize source data for cross-subject enhancement. Our uses parallel convolution capture temporal and spatial features first. Then, employ a novel attention-based adaptor that implicitly transfers domain, emphasizing global correlation features. We also use discriminator explicitly drive reduction marginal distribution discrepancy by against extractor adaptor. Besides, adaptive center loss is designed align conditional distribution. With aligned features, classifier optimized signals. Experiments on two widely datasets demonstrate our outperforms state-of-the-art methods, primarily due effectiveness These results indicate GAT has good potential enhance practicality BCI.

Language: Английский

---

Galea: A physiological sensing system for behavioral research in Virtual Environments

2020 IEEE Conference on Virtual Reality and 3D User Interfaces (VR), Journal Year: 2022, Volume and Issue: unknown, P. 66 - 76

Published: March 1, 2022

The pairing of Virtual Reality technology with Physiological Sensing has gained much interest in clinical settings and beyond: from developing novel methods for diagnosis perception cognition impairments, biofeedback anxiety treatment, to enhancing everyday practices such as self-guided meditation. However, conducting this type research does not come without challenges. For example, accessing the equipment recording data user synchronizing physiological response stimuli or interactive environment are trivial tasks, generating virtual content user's real-time is costly complex. This paper presents Galea, a device multi-modal signal acquisition able measure when experiencing content, enabling behavioral, affective computing , human-computer interaction applications access Parasympathetic nervous system Sympathetic simultaneously. We present primer on detectable human physiology an input source Computing perspective signals available through our device. describe primary design considerations circuit characterization results in-vivo recordings wearer's brain, eyes, heart, skin, muscles. also example help contextualize how these can be used reality setting. Galea makes working sensors more accessible offer standard inter intra experiment comparisons. Lastly, we discuss importance contributions work well future challenges that need considered.

Language: Английский

---

A Spectrally-Dense Encoding Method for Designing a High-Speed SSVEP-BCI With 120 Stimuli

IEEE Transactions on Neural Systems and Rehabilitation Engineering, Journal Year: 2022, Volume and Issue: 30, P. 2764 - 2772

Published: Jan. 1, 2022

The practical functionality of a brain-computer interface (BCI) is critically affected by the number stimuli, especially for steady-state visual evoked potential based BCI (SSVEP-BCI), which shows promise implementation multi-target system real-world applications. Joint frequency-phase modulation (JFPM) an effective and widely used method in modulating SSVEPs. However, ability JFPM to implement SSVEP-BCI with large e.g., over 100 remains unclear. To address this issue, spectrally-dense JPFM (sJFPM) proposed encode broad array modulates low- medium-frequency SSVEPs frequency interval 0.1 Hz triples stimuli conventional 120. validate effectiveness 120-target system, offline experiment subsequent online testing 18 healthy subjects total were conducted. verified feasibility using sJFPM designing 120 stimuli. Furthermore, demonstrated that achieved average performance 92.47±1.83% accuracy 213.23±6.60 bits/min information transfer rate (ITR), where more than 75% attained above 90% ITR 200 bits/min. This present study demonstrates elevating extends our understanding encoding means finer division.

Language: Английский

---

Cross-Subject Transfer Learning for Boosting Recognition Performance in SSVEP-Based BCIs

IEEE Transactions on Neural Systems and Rehabilitation Engineering, Journal Year: 2023, Volume and Issue: 31, P. 1574 - 1583

Published: Jan. 1, 2023

Steady-state visual evoked potential (SSVEP)-based brain-computer interfaces (BCIs) have been substantially studied in recent years due to their fast communication rate and high signal-to-noise ratio. The transfer learning is typically utilized improve the performance of SSVEP-based BCIs with auxiliary data from source domain. This study proposed an inter-subject method for enhancing SSVEP recognition through transferred templates spatial filters. In our method, filter was trained via multiple covariance maximization extract SSVEP-related information. relationships between training trial, individual template, artificially constructed reference are involved process. filters applied above form two new templates, obtained accordingly least-square regression. contribution scores different subjects can be calculated based on distance subject target subject. Finally, a four-dimensional feature vector detection. To demonstrate effectiveness publicly available dataset self-collected were employed evaluation. extensive experimental results validated feasibility improving

Language: Английский

---

Data Augmentation of SSVEPs Using Source Aliasing Matrix Estimation for Brain–Computer Interfaces

IEEE Transactions on Biomedical Engineering, Journal Year: 2022, Volume and Issue: 70(6), P. 1775 - 1785

Published: Dec. 7, 2022

Currently, ensemble task-related component analysis (eTRCA) and task discriminative (TDCA) are the state-of-the-art algorithms for steady-state visual evoked potential (SSVEP)-based brain-computer interfaces (BCIs). However, training BCIs requires multiple calibration trials. With insufficient data, accuracy of BCI will degrade, or even become invalid with only one trial. collecting a large amount electroencephalography (EEG) data is time-consuming laborious process, which hinders practical use eTRCA TDCA.This study proposed novel method, namely Source Aliasing Matrix Estimation (SAME), to augment SSVEP-BCIs. SAME could generate artificial EEG trials featured SSVEPs. Its effectiveness was evaluated using two public datasets (i.e., Benchmark, BETA).When combined SAME, both TDCA had significantly improved performance limited number data. Specifically, increased average by about 12% 3%, respectively, as few Notably, enabled work well single trial, achieving an >90% Benchmark dataset >70% BETA 1-second EEG.SAME effective method SSVEP-BCIs thereby enhancing TDCA.We propose new data-augmentation that compatible SSVEP-based BCIs. It can reduce efforts required calibrate SSVEP-BCIs, promising development

Language: Английский

---

Stimulus-Stimulus Transfer Based on Time-Frequency-Joint Representation in SSVEP-Based BCIs

IEEE Transactions on Biomedical Engineering, Journal Year: 2022, Volume and Issue: 70(2), P. 603 - 615

Published: Aug. 15, 2022

Brain-computer interfaces (BCIs) based on steady-state visual evoked potential (SSVEP) require extensive and costly calibration to achieve high performance. Using transfer learning re-use existing data from old stimuli is a promising strategy, but finding commonalities in the SSVEP signals across different remains challenge.This study presents new perspective, namely time-frequency-joint representation, which corresponding can be synchronized, thus emphasize common components. According this an adaptive decomposition technique multi-channel Fourier (MAFD) proposed adaptively decompose of simultaneously. Then, components identified transferred stimuli.A simulation public datasets demonstrates that stimulus-stimulus method has ability extract these stimuli. By using eight source stimuli, generate templates other 32 target It boosts ITR recognition 95.966 bits/min 123.684 bits/min.By extracting produces good classification performance without requiring stimuli.This provides synchronization standpoint analyze model signals. In addition, shortens time improve comfort, could facilitate real-world applications SSVEP-based BCIs.

Language: Английский

---

A Canonical Correlation Analysis-Based Transfer Learning Framework for Enhancing the Performance of SSVEP-Based BCIs

IEEE Transactions on Neural Systems and Rehabilitation Engineering, Journal Year: 2023, Volume and Issue: 31, P. 2809 - 2821

Published: Jan. 1, 2023

A steady-state visual evoked potential (SSVEP)- based brain-computer interface (BCI) can either achieve high classification accuracy in the case of sufficient training data or suppress stage at cost low accuracy. Although some researches attempted to conquer dilemma between performance and practicality, a highly effective approach has not yet been established. In this paper, we propose canonical correlation analysis (CCA)-based transfer learning framework for improving an SSVEP BCI reducing its calibration effort. Three spatial filters are optimized by CCA algorithm with intra- inter-subject EEG (IISCCA), two template signals estimated separately from target subject set source subjects six coefficients yielded testing signal each templates after they filtered three filters. The feature used is extracted sum squared multiplied their signs frequency recognized matching. To reduce individual discrepancy subjects, accuracy-based selection (ASS) developed screening those whose more similar subject. proposed ASS-IISCCA integrates both subject-specific models subject-independent information recognition signals. was evaluated on benchmark 35 compared state-of-the-art task-related component (TRCA). results show that significantly improve BCIs small number trials new user, thus helping facilitate applications real world.

Language: Английский

---

A new grid stimulus with subtle flicker perception for user-friendly SSVEP-based BCIs

Journal of Neural Engineering, Journal Year: 2023, Volume and Issue: 20(2), P. 026010 - 026010

Published: Feb. 24, 2023

Abstract Objective. The traditional uniform flickering stimulation pattern shows strong steady-state visual evoked potential (SSVEP) responses and poor user experience with intense flicker perception. To achieve a balance between performance comfort in SSVEP-based brain–computer interface (BCI) systems, this study proposed new grid reduced area low spatial contrast. Approach. A contrast scanning experiment was conducted first to clarify the relationship SSVEP characteristics signs values of Four patterns were involved experiment: ON OFF that separately activated positive or negative information processing pathways, ON–OFF simultaneously both served as control group. contrast-intensity contrast-user curves obtained for each pattern. Accordingly, optimized schemes (the ON-50% stimulus, OFF-50% Flicker-30% stimulus) applied 12-target 40-target BCI speller compared stimulus Flicker-500% evaluation subjective experience. Main results. showed comparable online (12-target, 2 s: 69.87 ± 0.74 vs. 69.76 0.58 bits min −1 , 40-target, 4 57.02 2.53 60.79 1.08 ) improved (better comfortable level, weaker perception higher preference level) multi-targets spellers. Significance. Selective activation pathway using robust responses. On basis, high-performance user-friendly BCIs have been developed implemented, which has important theoretical significance application value promoting development technology.

Language: Английский

---