Computer Science Department, PhD Proposal Defense - Ruixue Liu " Decoding Cognitive States from fNIRS Data Using Machine Learning"

Monday, December 16, 2019
12:00 pm to 1:00 pm
Floor/Room #: 

Ruixue Liu

PhD Student

WPI – Computer Science

Committee members:

Prof. Erin Solovey, WPI - Computer Science (Advisor)

Prof. Rodica Neamtu, WPI - Computer Science

Prof. Ali Yousefi, WPI - Computer Science

Prof. Kate Arrington, Lehigh University (External Member)


 Automatic detection of an individual's cognitive states has implications in many domains, such as gaming, driving, and learning. However, cognitive states have been challenging to identify using traditional measures. Brain-based sensing offers an alternative by tracking internal states through measuring brain activity. Recently, the use of functional near-infrared spectroscopy (fNIRS) has received focus because of its promise for detecting an individual's cognitive state in more ecologically valid studies.

The primary goal for this dissertation is to investigate the feasibility of decoding different cognitive states from fNIRS data, through developing and applying novel machine learning methods. We approach this goal through three primary research tasks:

(1) Developing a machine learning framework to differentiate an individual's mind-wandering state versus on-task state using fNIRS. Specifically, we conduct a study using fNIRS during the Sustained Attention to Response Task (SART) task to elicit mind-wandering states. We then build machine learning classifiers both on an individual level and a group level. We also propose an individual-based novel window selection algorithm to improve classification accuracy.

(2) Exploring the feasibility of detecting different cognitive states based on fNIRS data collected from multiple studies. Specifically, we conduct a second study using fNIRS during a rule-learning task. We then combine the data collected during the rule-learning task and the SART task, to classify the learning state versus mind-wandering state from the two studies.  We aim to apply advanced deep learning models to achieve state-of-the-art classification accuracy.

(3) Investigating novel machine learning frameworks for classifying drivers' mental workload using fNIRS. We conduct an experiment using fNIRS in a simulated driving environment, where participants were asked to perform a driving task and the n-back task. The n-back task can elicit different levels of mental workload. We investigate how to apply reservoir computing frameworks to improve the classification accuracy, as well as reduce the computational cost.