BCI

Our research group focuses on advancing Brain-Computer Interface (BCI) technology—an innovative system enabling direct communication between the brain and computers.

At the core of our research are electroencephalogram (EEG) signals, which we process through cutting-edge methods in preprocessing, feature extraction, feature selection, and classification. These methods are critical for improving the accuracy and reliability of BCI systems.

Our goal is to develop a real-time, highly accurate, and physiologically interpretable asynchronous BCI system. To achieve this, we are pursuing three objectives:

Designing a feature selection algorithm with physiological interpretability.
Creating a nonlinear graph-embedding-based feature extraction algorithm considering spatial relationships.
Adapting change-point detection algorithms to identify mental tasks in EEG signals.

By addressing these objectives, we aim to reduce the complexity of EEG data, both in sample size and feature dimensions, enabling faster and more accurate classification models. The performance of our asynchronous BCI system will be rigorously tested with real-world data collected through the developed platform.

Interpretable Feature Extraction And Selection for Time Series

The secondary objective of our lab is to develop an interpretable framework for evaluating sensor data collected from industrial environments. With the advent of affordable Industrial Internet of Things (IIoT) devices, data collection has become relatively inexpensive. However, using this data effectively—whether for forecasting, modeling, or analysis—within an interpretable framework remains a significant challenge. The BCI-Feast project aims to address this issue by designing innovative pre- and post-processing mechanisms to enhance the physical interpretability and utility of the data.

To achieve this, the team focuses on advanced methodologies, including nonlinear time series analysis and change-point detection algorithms, to mitigate the inherent non-stationary behavior and noise present in industrial data. Preliminary results from the ongoing research provide valuable insights and direction for further development.

The overarching vision of the lab can be encapsulated as: "Bringing Lost Data (in Space and Time) Back to Reality." This vision is being realized through a multidisciplinary approach that integrates linear algebra, statistics, and computer science, driven by intelligence—not the artificial one.