Analyzing time series from eye tracking using Symbolic Aggregate Approximation

Abstract

This thesis explores the viability of transforming the data produced when tracking the eyes into a discrete symbolic representation. For this transformation, we utilize Symbolic Aggregate Approximation to investigate a new possibility for effectively categorizing data collected via eye tracking technologies. This categorization illustrates tendencies for, e.g., tracking problems, problems with the set-up, normal vision, or vision disturbances. Accordingly, this will contribute to evaluating the eyes' performance and allow professionals to develop a diagnosis based on evidence from objective measurements. The results are based on implementing a symbolic discretization method applied to experiments on a real-world dataset containing recordings of eye movements. In the future, the knowledge and transformation via the SAX method can be utilized to make sense of data and identify anomalies implemented in various domains and for multiple stakeholders.