A survey on encoding and decoding technology of non-invasive brain-computer interface

Shuang Qiu, Banghua Yang, Xiaogang Chen, Yijun Wang, Minpeng Xu, Baoliang Lyu, Xiaorong Gao, Huiguang HeJournal of Image and Graphics 2023 · DOI

Abstract

脑—机接口(brain-computer interface,BCI)系统通过采集、分析大脑信号,将其转换为输出指令,从而跨越外周神经系统,实现由大脑信号对外部设备的直接控制,进而用于替代、修复、增强、补充或改善中枢神经系统的正常输出。非侵入式脑—机接口由于具有安全性以及便携性等优点,得到了广泛关注和持续研究。研究人员对脑信号编码方法的不断探索扩展了BCI系统的应用场景和适用范围。同时,脑信号解码方法的不断研发极大地克服了脑电信号信噪比低的缺点,提高了系统性能,这都为构建高性能脑—机接口系统奠定了基础。本文综述了非侵入式脑—机接口编解码技术以及系统应用的最新研究进展,展望其未来发展前景,以期促进BCI系统的深入研究与广泛应用。

Introduction: The Challenge of Non-Invasive BCI

Brain-Computer Interfaces (BCIs) establish direct communication between the brain and external devices, bypassing traditional neuromuscular pathways. While invasive BCIs offer high signal quality, they involve surgical risks. Non-invasive BCIs, primarily utilizing Electroencephalography (EEG), are safer and more portable but face significant hurdles: low signal-to-noise ratio (SNR), high inter-subject variability, and the need for lengthy calibration sessions. Recent research focuses on advanced encoding paradigms and decoding algorithms to overcome these barriers.

General architecture of a non-invasive BCI system, illustrating the flow from signal acquisition to application feedback.

Encoding Paradigms: From Coarse to Fine Control

The effectiveness of a BCI depends on how brain signals are triggered (encoding). Traditional paradigms are evolving to support more complex interactions:

  1. Motor Imagery (MI): Users imagine limb movements. Research is shifting from simple binary tasks (e.g., left vs. right hand) to "fine" motor tasks involving specific joints and multi-directional trajectories.
  2. Steady-State Visual Evoked Potential (SSVEP): Triggered by flickering stimuli. Modern decoding has significantly increased the information transfer rate (ITR), with some systems reaching over 150 bit/min.
  3. Event-Related Potential (ERP): Utilizes the P300 wave. Hybrid systems (e.g., SSVEP + P300) are now used to expand command sets, enabling high-speed spellers with over 100 characters.
  4. Affective BCI: Detects emotional states, often using multimodal fusion (EEG combined with eye-tracking or physiological signals) to improve accuracy.
Illustration of fine motor imagery tasks, such as multi-directional arm reaching, which expand the control capabilities of BCIs.

Decoding Innovations: Deep Learning and Transfer Learning

The core of modern BCI progress lies in decoding—interpreting noisy EEG signals. The field has moved from traditional spatial filtering (like CSP) to Deep Learning architectures such as EEGNet, Transformers, and CNNs. These models are better at capturing complex temporal and spatial features of brain activity.

Furthermore, Transfer Learning and Domain Adaptation are addressing the "calibration fatigue" problem. By leveraging data from multiple subjects, researchers are developing "zero-calibration" or "plug-and-play" systems that generalize across different users without requiring extensive individual training.

Key Insights and Clinical Impact

  • Hybridization: Combining different paradigms allows for high-performance systems that outperform single-modality approaches.
  • Zero-Calibration: Meta-learning and adversarial training are paving the way for BCIs that work immediately for new users.
  • Rehabilitation: MI-BCIs combined with functional electrical stimulation (FES) are showing promise in stroke rehabilitation by promoting neuroplasticity.

Future Directions

Despite progress, issues like "BCI illiteracy" (where some users cannot generate usable signals) and visual fatigue from SSVEP stimuli remain. Future research will likely integrate 5G/6G and cloud computing to handle the computational demands of deep learning in real-time, alongside deeper human-AI collaboration to create more robust, "in-the-wild" BCI applications.

Reference: Qiu, S., et al. (2023). A survey on encoding and decoding technology of non-invasive brain-computer interface. Journal of Image and Graphics, 28(06): 1543-1566. DOI: 10.11834/jig.230031.

Summary updated

Abstract

脑—机接口(brain-computer interface,BCI)系统通过采集、分析大脑信号,将其转换为输出指令,从而跨越外周神经系统,实现由大脑信号对外部设备的直接控制,进而用于替代、修复、增强、补充或改善中枢神经系统的正常输出。非侵入式脑—机接口由于具有安全性以及便携性等优点,得到了广泛关注和持续研究。研究人员对脑信号编码方法的不断探索扩展了BCI系统的应用场景和适用范围。同时,脑信号解码方法的不断研发极大地克服了脑电信号信噪比低的缺点,提高了系统性能,这都为构建高性能脑—机接口系统奠定了基础。本文综述了非侵入式脑—机接口编解码技术以及系统应用的最新研究进展,展望其未来发展前景,以期促进BCI系统的深入研究与广泛应用。
BibTeX
@article{qiu2023survey,
  author = {Shuang Qiu and Banghua Yang and Xiaogang Chen and Yijun Wang and Minpeng Xu and Baoliang Lyu and Xiaorong Gao and Huiguang He},
  title = {{A survey on encoding and decoding technology of non-invasive brain-computer interface}},
  journal = {Journal of Image and Graphics},
  year = {2023},
  doi = {10.11834/jig.230031},
}
Figure