Toward Real-time BCI data analysis with Deep Neural Networks

Abstract

Recent work has demonstrated that Functional Near-Infrared Spectroscopy (fNIRS) is potentially more suitable (than EEG) for Brain-Computer Interaction. Meanwhile, it has been proved that deep neural networks can achieve state-of-the-art results on lots of classic machine learning tasks. However, it is not clear we can achieve the same success in building a better Brain-Computer Interface (BCI).

We applied the deep learning models to two small-sized fNIRS datasets. We were then inspired to design and develop an automated framework aiming to build a high quality public large-scale fNIRS dataset. At last, we will discuss selection, optimization, and interpretation of existing models and future work/challenges while implementing a real-time BCI framework.

Join meeting in Sococo VH 209. Login: tuftscs.sococo.com.

Join Zoom Meeting: https://tufts.zoom.us/j/98610939077

PASSWORD: See colloquia email

Dial by our location: +1 646 558 8656 US (New York)

Passcode: See colloquia email