A probabilistic real-time algorithm for detecting blinks, saccades, and fixations from EOG data

Miika Toivanen, Kati Pettersson, Kristian Lukander

Abstract


We present a computationally light real-time algorithm which automatically detects blinks, saccades, and fixations from electro-oculography (EOG) data and calculates their temporal parameters. The method is probabilistic which allows to consider the uncertainties in the detected events. The method is real-time in the sense that it processes the data sample-by-sample, without a need to process the whole data as a batch. Prior to the actual measurements, a short, unsupervised training period is required. The parameters of the Gaussian likelihoods are learnt using an expectation maximization algorithm. The results show the promise of the method in detecting blinks, saccades, and fixations, with detection rates close to 100 %. The presented method is published as an open source tool.

Keywords


electro-oculography; blinks; saccades; fixations; probabilistic modeling; algorithms

Full Text:

PDF


DOI: http://dx.doi.org/10.16910/jemr.8.2.1

Refbacks

  • There are currently no refbacks.


Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 International License.