Taming fNIRS-based BCI Input for Better Calibration and Broader Use

Currently, we are waiting for the Tufts IRB to give us permission to release the participants' data. (We are almost there, and we apologize for the inconvenience).

If you want to download the data or have any questions, please reach out to Leon (leonwang_at_cs.tufts.edu).

# Paper Information

L. Wang*, Z. Huang*, Z. Zhou, D.T. McKeon, G. Blaney, M.C. Hughes†, and R.J.K Jacob†, “Taming fNIRS-based BCI Input for Better Calibration and Broader Use,” Proc. ACM UIST 2021 Symposium on User Interface Software and Technology (2021).

*Both authors contributed equally to this research.

†Both authors jointly supervised this work.

# Description

We have presented new tools and a new dataset intended to allow future researchers and designers to create and explore fNIRS-based BCI applications more easily. Some examples of the implications for the design of this work are found in such previous experimental fNIRS-based systems as: Brainput for robot automation, air traffic control, music learning, and bubble cursor usage. We provide a new fNIRS dataset collected using rigorous procedures from subjects performing a standard n-back cognitive task and show how it can be used to improve performance for future systems. We developed a new machine learning approach to process and utilize fNIRS data. We show from our experiments that our proposed 3-phase machine learning pipeline significantly improves n-back task classification performance over several established baselines. Moreover, even with a reduced amount of per-user training data, our approach still outperforms baseline models trained with all available target-subject-specific training data, showing the potential of reducing individual calibration effort when deploying BCI applications in the future. We hope that our new dataset, machine learning method, and tools will remove barriers that have prevented a wider range of researchers from using fNIRS-based BCI and facilitate the development of a new generation of powerful and easy use brain-computer interfaces.

# Code

Here is the link to our code Github repo (opens new window)

# Dataset

Paper link (opens new window)

# Bibtex

  title = {Taming {{fNIRS}}-Based {{BCI Input}} for {{Better Calibration}} and {{Broader Use}}},
  booktitle = {The 34th {{Annual ACM Symposium}} on {{User Interface Software}} and {{Technology}} ({{UIST}} '21)},
  author = {Wang, Liang and Huang, Zhe and Zhou, Ziyu and McKeon, Devon and Blaney, Giles and Hughes, Michael C and Jacob, Robert J. K.},
  year = {2021},
  url = {https://www.cs.tufts.edu/~jacob/papers/uist21.pdf}
The Tufts fNIRS to Mental Workload Dataset

The Tufts fNIRS to Mental Workload Dataset