BCILAB is conducting neural data science researches on reading out neural information for understanding brain mechanisms and developing brain-computer interfaces (BCIs).

 

Using AI-powered data science tools, we aim to understand how the brain works by cracking a wide array of brain signals, including neuronal spikes, local field potentials (LFP), Calcium imaging, electrocorticography (ECoG), electroencephalography (EEG), magnetoencephalography (MEG), and functional magnetic resonance imaging (fMRI). By doing this research, we develop computational methods to understand neural data as well as computational models to explain neural computation. Ultimately, we aim to translate our understanding to clinical and engineering applications.

 

We are building both non-invasive and invasive BCIs. Non-invasive BCIs built in our lab provide a new user interface to control home appliances or other devices using event-related potentials (ERPs) embedded in the human EEG. We aim to make our EEG BCIs available for daily use at home by maximizing performance to a practical level and circumventing real-world hurdles such as individual variations, daily calibration and cognitive distractions, to name a few. In addition, we are developing a bi-directional BCI to control external devices by reading motor cortical activities and at the same time, to deliver somatosensory senses back to the brain by writing the code of senses directly to the brain. With an invasive BCI, we hope to provide independence and enhanced quality of life to severely disabled persons.

  

Additionally, we study tactile information processing. First, we build artificial neural networks (ANNs) by mimicking the structure and function of human somatosensory system. The ANNs are designed to learn tactile percepts from artificial tactile sensor signals. Next, we engineer electrical stimulation patterns to generate diverse tactile senses in humans. We believe that these non-invasive peripheral electrical stimulators will offer a novel haptic interface.