eduzhai > Applied Sciences > Engineering >

Real-Time Face and Landmark Localization for Eyeblink Detection

  • Save

... pages left unread,continue reading

Document pages: 13 pages

Abstract: Pavlovian eyeblink conditioning is a powerful experiment used in the field ofneuroscience to measure multiple aspects of how we learn in our daily life. Totrack the movement of the eyelid during an experiment, researchers havetraditionally made use of potentiometers or electromyography. More recently,the use of computer vision and image processing alleviated the need for thesetechniques but currently employed methods require human intervention and arenot fast enough to enable real-time processing. In this work, a face- andlandmark-detection algorithm have been carefully combined in order to providefully automated eyelid tracking, and have further been accelerated to make thefirst crucial step towards online, closed-loop experiments. Such experimentshave not been achieved so far and are expected to offer significant insights inthe workings of neurological and psychiatric disorders. Based on an extensiveliterature search, various different algorithms for face detection and landmarkdetection have been analyzed and evaluated. Two algorithms were identified asmost suitable for eyelid detection: the Histogram-of-Oriented-Gradients (HOG)algorithm for face detection and the Ensemble-of-Regression-Trees (ERT)algorithm for landmark detection. These two algorithms have been accelerated onGPU and CPU, achieving speedups of 1,753$ times$ and 11$ times$, respectively.To demonstrate the usefulness of our eyelid-detection algorithm, a researchhypothesis was formed and a well-established neuroscientific experiment wasemployed: eyeblink detection. Our experimental evaluation reveals an overallapplication runtime of 0.533 ms per frame, which is 1,101$ times$ faster thanthe sequential implementation and well within the real-time requirements ofeyeblink conditioning in humans, i.e. faster than 500 frames per second.

Please select stars to rate!


0 comments Sign in to leave a comment.

    Data loading, please wait...