Unsupervised parsing of gaze data with a beta-process vector auto-regressive hidden Markov model

The first stage of analyzing eye-tracking data is commonly to code the data into sequences of fixations and saccades. This process is usually automated using simple, predetermined rules for classifying ranges of the time series into events, such as "if the dispersion of gaze samples is lower than a particular threshold, then code as a fixation; otherwise code as a saccade." More recent approaches incorporate additional eye-movement categories in automated parsing algorithms by using time-varying, data-driven thresholds. We describe an alternative approach using the beta-process vector auto-regressive hidden Markov model (BP-AR-HMM). The BP-AR-HMM offers two main advantages over existing frameworks. First, it provides a statistical model for eye-movement classification rather than a single estimate. Second, the BP-AR-HMM uses a latent process to model the number and nature of the types of eye movements and hence is not constrained to predetermined categories. We applied the BP-AR-HMM both to high-sampling rate gaze data from Andersson et al. (Behavior Research Methods 49(2), 1-22 2016) and to low-sampling rate data from the DIEM project (Mital et al., Cognitive Computation 3(1), 5-24 2011). Driven by the data properties, the BP-AR-HMM identified over five categories of movements, some which clearly mapped on to fixations and saccades, and others potentially captured post-saccadic oscillations, smooth pursuit, and various recording errors. The BP-AR-HMM serves as an effective algorithm for data-driven event parsing alone or as an initial step in exploring the characteristics of gaze data sets.