WearBCI Dataset: Understanding and Benchmarking Real-World Wearable Brain-Computer Interfaces Signals

Haoxian Liu; Hengle Jiang; Lanxuan Hong; Xiaomin Ouyang

cs.HCMar 29, 2026

WearBCI Dataset: Understanding and Benchmarking Real-World Wearable Brain-Computer Interfaces Signals

Authors: Haoxian Liu, Hengle Jiang, Lanxuan Hong, Xiaomin Ouyang

Organizations: Hong Kong University of Science and Technology, Hong Kong SAR, China

Abstract

Brain-computer interfaces (BCIs) have opened new platforms for human-computer interaction, medical diagnostics, and neurorehabilitation. Wearable BCI systems, which typically employ non-invasive electrodes for portable monitoring, hold great promise for real-world applications, but also face significant challenges of signal quality degradation caused by motion artifacts and environmental interferences. Most existing wearable BCI datasets are collected under stationary or controlled lab settings, limiting their utility for evaluating performance under body movement. To bridge this gap, we introduce WearBCI, the first dataset that comprehensively evaluates wearable BCI signals under different motion dynamics with synchronized multimodal recordings (EEG, IMU, and egocentric video), and systematic benchmark evaluations for studying impacts of motion artifact. Specifically, we collect data from 36 participants across different motion dynamics, including body movements, walking, and navigation. This dataset includes synchronized electroencephalography (EEG), inertial measurement unit (IMU) data, and egocentric video recordings. We analyze the collected wearable EEG signals to understand the impact of motion artifacts across different conditions, and benchmark representative EEG signal enhancement techniques on our dataset. Furthermore, we explore two new case studies: cross-modal EEG signal enhancement and multi-dimension human behavior understanding. These findings offer valuable insights into real-world wearable BCI deployment and new applications.

Figures & tables

Figure 1. Comparison of traditional and wearable BCIs.

Panel of Figure 2 — Figure 2. Overview of the WearBCI experimental setting and recorded multimodal data.

Subjects	36 healthy adults (18 - 26 yrs), balanced gender ratio
Total Duration	EEG: 16.4 h, IMU: 9.2 h, Video: 0.9 h
Modalities	8-ch EEG (250 Hz), 5 IMUs (100 Hz), egocentric video (30 FPS)
Sessions	Static, Body Movements (5 types), Walking (3 speeds), Navigation

Table 1. Summary of the WearBCI Dataset.

Panel of Figure 3 — Figure 3. Time-domain waveforms and motion-EEG correlation across different motion levels.

Panel of Figure 4 — Figure 4. PSD of EEG signals in different settings.

Figure 5. Topographic maps of different brain regions. Colors indicate PSD values, blue is higher.

Method

Body Movements

Walking

Navigation

D_{\mathrm{PSD}}

D_{\alpha}

EMG Index

Brain

D_{\mathrm{PSD}}

D_{\alpha}

EMG Index

Brain

D_{\mathrm{PSD}}

D_{\alpha}

EMG Index

Brain

ICA

16.1\pm 15.5

2.1\pm 0.4

1.3\pm 1.8

\mathbf{0.47\pm 0.19}

31.5\pm 16.2

\mathbf{1.6\pm 0.4}

1.1\pm 9.3

0.30\pm 0.17

50.6\pm 34.9

\mathbf{1.7\pm 0.6}

1.7\pm 0.4

0.50\pm 0.35

EMD

25.0\pm 19.4

\mathbf{1.7\pm 0.3}

\mathbf{0.2\pm 0.1}

0.38\pm 0.15

96.4\pm 40.3

3.4\pm 0.4

\mathbf{0.3\pm 0.2}

0.38\pm 0.18

213.9\pm 167.9

1.9\pm 0.4

\mathbf{0.5\pm 0.1}

0.41\pm 0.12

ASR

12.8\pm 12.6

2.1\pm 0.5

1.2\pm 0.7

0.28\pm 0.21

23.6\pm 7.5

2.2\pm 0.7

2.3\pm 2.9

\mathbf{0.45\pm 0.23}

145.0\pm 155.1

1.8\pm 0.3

1.6\pm 0.2

\mathbf{0.66\pm 0.16}

GCTNet

13.7\pm 11.2

2.3\pm 0.3

2.0\pm 0.4

0.20\pm 0.10

16.4\pm 62.8

2.1\pm 0.4

2.0\pm 0.4

0.33\pm 0.20

9.0\pm 3.5

2.4\pm 0.2

1.9\pm 0.1

0.19\pm 0.22

EEGDnet

9.5\pm 3.2

2.4\pm 0.1

1.3\pm 0.3

0.17\pm 0.07

10.2\pm 1.9

2.3\pm 0.3

1.3\pm 0.1

0.25\pm 0.05

9.6\pm 3.4

2.4\pm 0.2

1.2\pm 0.2

0.28\pm 0.16

Table 2. Benchmarking EEG denoising methods across three motion conditions relative to a static intra-subject eyes-closed setting. The results reveal performance degradation under increasing motion and highlight trade-offs between artifact suppression and neural preservation. Bold denotes best , underline denotes second best in each column and motion condition.

Figure 6. Analysis of noise reduction effects in different scenes of navigation.

Figure 7. IMU-assisted EEG signal enhancement.

Panel of Figure 8 — Figure 8. Comparison of PSD and waveform before (R: Raw EEG) and after (C: Clean EEG) IMU-assisted signal enhancement during walking.

Figure 9. Illustration of multi-dimension behavior understanding through different modalities.

Modality	Cognition	Scene	Action	Overall
EEG-only	0.451	0.271	0.573	0.386
IMU-only	0.168	0.143	0.649	0.330
Video-only	0.276	0.864	0.652	0.622
Concatenate Fusion	0.477	0.885	0.229	0.727
Optimized Fusion	0.413	0.907	0.729	0.859

Table 3. Quantitative evaluation of semantic alignment (BERTScore F 1 ) across modalities and fusion strategies.

EEG Artifact Removal

See all 10 papers

Synthetic Data Generation for Brain-Computer Interfaces: Overview, Benchmarking, and Future Directions
Mar 11, 2026Ziwei Wang, Zhentao He, Xingyi He +6Synthetic Data Generation Brain-Computer Interfaces
School of Artificial Intelligence and Automation, Huazhong University of Science and Technology, Wuhan, China · Zhongguancun Academy, Beijing, China
EgoInertia-MI: A Multimodal Egocentric Vision and IMU Benchmark for Motor Impairment Assessment
Jul 4, 2026Fatemah Alhamdoosh, Pietro Pala, Abduallah Mohamed +1Human Activity Recognition Wearable Sensing
University of Florence, Italy · Meta Reality Labs, USA · University of Edinburgh, UK
NeuroPath: Practically Adopting Motor Imagery Decoding through EEG Signals
Mar 30, 2026Jiani Cao, Kun Wang, Yang Liu +1MI Classification Brain-Computer Interfaces
Department of Computer Science City University of Hong Kong Hong Kong, China · Department of Computer Science Florida State University Tallahassee, United States

Mar 11, 2026J/K move · Enter open · S save

Synthetic Data Generation for Brain-Computer Interfaces: Overview, Benchmarking, and Future Directions

Ziwei Wang, Zhentao He, Xingyi He, Hongbin Wang+5

School of Artificial Intelligence and Automation, Huazhong University of Science and Technology, Wuhan, China · Zhongguancun Academy, Beijing, China

Synthetic Data Generation Brain-Computer Interfaces Synthetic Data Evaluation Benchmark Design Electroencephalography

Open paper AskPDF arXiv

Deep learning has achieved transformative performance across diverse domains, largely driven by large-scale and high-quality training data. In contrast, the development of brain-computer interfaces (BCIs) is fundamentally constrained by limited, heterogeneous, and privacy-sensitive neural recordings. Generating synthetic yet physiologically plausible brain signals has therefore emerged as a promising strategy to mitigate data scarcity, improve model generalization, and support data-efficient BCIs. This survey provides a comprehensive review of synthetic brain data generation for BCIs, covering methodological taxonomies, benchmark experiments, evaluation metrics, key applications, and future directions. We systematically categorize existing generation approaches into four types: signal-transformation-based, feature-based, model-based, and translation-based generation, and discuss their characteristics, advantages, and limitations. Furthermore, we benchmark representative brain signal generation approaches across four BCI paradigms, including motor imagery, epileptic seizure detection, steady-state visually evoked potentials, and auditory attention detection, to provide an objective comparison of their downstream utility. We also summarize evaluation principles for generated brain signals from multiple perspectives, including signal realism, physiological plausibility, downstream utility, and privacy preservation. Finally, we discuss the potential and challenges of current generation approaches and outline future research directions toward accurate, data-efficient, generalizable, and privacy-aware BCI systems. The benchmark codebase is available at https://github.com/wzwvv/DG4BCI.

Abstract

EEG Artifact Removal

Explore similar work

Synthetic Data Generation for Brain-Computer Interfaces: Overview, Benchmarking, and Future Directions

EgoInertia-MI: A Multimodal Egocentric Vision and IMU Benchmark for Motor Impairment Assessment

NeuroPath: Practically Adopting Motor Imagery Decoding through EEG Signals