DISCUSSION

1. EYE FIXATIONS IDENTIFICATION BASED ON STATISTICAL ANALYSIS - CASE STUDY Giacomo Veneri, Pietro Piu, Pamela Federighi,Francesca Rosini, Antonio Federico, Alessandra Rufa Eye tracking & Vision Applications Lab (EVA Lab), Section Clinical Neurophysiology, Department of Neurological Neurosurgical and Behavioral Science, University of Siena, Italy Etruria Innovazione scpa - via Banchi di Sotto, 55, 53100 Siena, Italy • Objectives • Eye movement is the most simple and repetitive movement that enable humans to interact with the environment. The common daily activities, such as watching television or reading a book, involve this natural activity which consists of rapidly shifting our gaze from one region to another. The identification of the main components of eye movement during visual exploration such as fixations and saccades, is the objective of the analysis of eye movements in various contexts ranging from basic neuro sciences and visual sciences to virtual reality interactions and robotics. However, many of the algorithms that detect fixations present a number of problems. In this article, we present a new fixation identification algorithm based on the analysis of variance and F-test. We present the new algorithm and we compare it with the common fixations algorithm based on dispersion. To demonstrate the performance of our approach we tested the algorithm in a group of healthy subjects. Background From a neurological point of view, fixation is defined as the act to maintain the visual gaze on a single location in order to make our environment visible. From a technical point of view, fixation should be identified by a cluster of point around a centroid with a minimum duration. During fixation, the eye does not remain completely perturbations, but is affected by perturbations such as microsaccades, ocular drifts, and ocular micro tremor, making it difficult to easily identify by an algorithm. INTRODUCTION Small portion of gaze of subject GV on a free visual search task (balloon image). Image converted to grayscale only for readable purpose. METHODS Healthy subjects experiment design Five subjects were enrolled aged 25-45. Subjects were seated at viewing distance of 78cm from a 32'' color monitor (51cmx31cm). Eye position was recorded using ASL 6000 system, which consists of a remote-mounted camera sampling pupil location at 240Hz. A 9-point calibration and 3-point validation procedure was repeated several times to ensure all recordings had a mean spatial error of less than 0.3 degree. Data was controlled by a Pentium4 dual core 3GHz computer, which acquires signals by fast UART serial port. Head movement was restricted using a chin rest and bite. Subject was asked to fixate a centered red dot on the center; after 500ms move the centered red dot disappear and subject can explore the scene. Displayed scenes were randomly choose in a collection of real image or pop up image. ImplemenetedMethod The key principle of the proposed technique is based in supposing the variance along the x is not significantly different from that along the y during a fixations. The algorithm is based on this assumption, and tests the hypothesis by a statistical method such as the F-test. Algorithm RESULTS FDT versus I-DT Tableshowsfixations identified by algorithms ftd and i-dt on five healthy subjects. Graphs of 3 healthy subjects. Percentage of correct fixation identified by I-DT and FDT varying variance of x axis, y axis and duration of fixation. Performance of I-DT and FDT are similar for small value of σx and σy; when dispersion of x and/or y increased I-DT was not able to identify fixation as expected, on the contrary, performance of FDT remained stable. Performance of FDT appeared to decrease when σx and σy were different. DISCUSSION The FDT identified simulated fixations with a precision of the 87% and I-DT the 42%. The performance of I-DT, however, strictly depends on the threshold dispersion set; when we chose a large threshold the I-DT identified more fixations, but in a real case should be affected by incorrect segmentation of saccade. FDT failed when it identified one fixation as two fixations(over segmentation problem). Due to the previous issues it is not easy to compare I-DT and FDT. We suggested a formal definition of fixations based on analysis of variance between x axis and y axis; the implemented algorithm is based on the dispersion algorithm I-DT developed by Salvucci and Goldberg and integrates it with a statistical test (F-test). The main advantage of the proposed technique is to provide a new definition of fixation which does not require the setting of any critical parameter or threshold, and provides a probability value of correctness. www: http://www.evalab.unisi.it – email: visionlab@unisi.it – skype: eyetracking www: http://www.etinnova.it – email: giacomo.veneri#etinnova.it CONCLUSION