Thema
Local Condensing of Gaze Samples for Clustering-Based Smooth Pursuit Detection
Typ Master, Forschungspraxis, Bachelor, Ing.prax.
Betreuer Mikhail Startsev
Tel.: +49 (0)89 289-28550
E-Mail: mikhail.startsev@tum.de
Sachgebiet Human and Computer Vision
Beschreibung In order to keep track of their environment, humans keep moving their eyes several times a second. Humans exhibit multiple types of eye movements, which are in different ways responsible for perception. Precise separation of these movements from one another is therefore desirable.

In laboratory conditions image and video stimuli are highly prevalent. While for images we observe two major eye movement types (fixations, relatively stationary, and saccades, characterized by very high speed and acceleration), for videos we can add at least one more eye movement type: the smooth pursuit (sometimes called also a “fixation on a moving target”), where the gaze follows some object or otherwise “interesting” (salient) patch on the scene in a relatively smooth fashion. This particular movement of the eye is fairly hard to detect, ex. due to the interference of the noise during eye tracking recording or the fact that the eye is not in a strict sense motionless during fixations, so a slow pursuit motion can be confused with a noisy fixation.

To counter the shortcomings of the previously existing detection methods, we developed an algorithm that utilizes the recordings of eye movements for multiple observers, clustering them to detect common motion patterns. Such an approach can benefit from a preprocessing step that could locally condense the gaze samples (whilst preserving the relation to the original data points), for example to the points of local extrema, like several steps of a Mean-Shift Algorithm would. The candidate would implement this idea in combination with an existing framework for smooth pursuit detection [1].



[1] Agtzidis, I., Startsev, M., and Dorr, M. 2016. Smooth Pursuit Detection Based on Multiple Observers. ETRA 2016
Voraussetzung Knowledge of Python is desirable, some algorithmic and/or machine learning background is a plus.
Bewerbung If you are interested in this topic, we welcome the applications via the email address above. Please set the email subject to “<Type of application> application for topic 'XYZ'”, ex. “Master’s thesis application for topic 'XYZ'”, while clearly specifying why are you interested in the topic in the text of the message. Also make sure to attach your most recent CV (if you have one) and grade report.