Date of Award


Document Type

Campus Access Dissertation

Degree Name

Doctor of Philosophy (PhD)


Computer Science

First Advisor

Marc Pomplun

Second Advisor

Dan Simovici

Third Advisor

Ming Ouyang


Eye movements are any shifts in position of the eye in its orbit, serving as an important mechanism for humans to select information from their environment. Eye movements can be recorded by an eye tracker, which measures where the eyes are focused, how we move the gaze, and how the pupils react to different stimuli. This information associated with eye movements enables a fine-grained analysis of human behaviors that can be applied in a wide variety of fields, from academic and scientific research, such as psychology, neuroscience and medicine, to practical applications, such as human-computer interfaces, advanced driver assistance systems and biometrics.

In the first part of the dissertation, I present a general recurrent neural network (RNN) framework for biometric recognition based on eye movements. Unlike the traditional eye movement biometrics that employ handcrafted features which lead to complex computation and heavy reliance on experimental design, the proposed model can automatically learn the dynamic features and temporal dependencies from a narrow data window extracted from a sequence of raw eye movement signals. I evaluate the model on a dataset with 32 subjects presented with static images, and the results show that the model significantly outperforms previous methods. The second part of the dissertation studies in-vehicle eye tracking accuracy maintenance. I propose an approach to automatically perform real-time eye tracking calibration. I apply the Integral Channel Features approach to continually detect traffic signs that would likely attract the driver's attention, which in turn are used as moving stimuli for real-time eye tracking calibration. The error vectors between the recorded fixations and the moving targets are calculated immediately and the weighted average of them is used to compensate for the offsets of fixations in real-time. I evaluate the method both on laboratory data and real driving data and show that it can effectively reduce the measurement errors.


Free and open access to this Campus Access Dissertation is made available to the UMass Boston community by ScholarWorks at UMass Boston. Those not on campus and those without a UMass Boston campus username and password may gain access to this dissertation through resources like Proquest Dissertations & Theses Global or through Interlibrary Loan. If you have a UMass Boston campus username and password and would like to download this work from off-campus, click on the "Off-Campus UMass Boston Users" link above.