Infant Gaze and Head Tracking in a Wide Visual Field: Advantages and Challenges

• In Event: Poster Session 07
  In Poster Session: PS 07 Topic - Methods, History, Theory

Abstract

In the past decades, eye tracking techniques have provided a simple and accessible tool to measure infant’s looking behaviour and gain insights into early perception and cognition (see Gredebäck, Johnson & von Hofsten, 2009, for a review). The most well-established eye tracking paradigms in infancy have taken advantage of remote eye tracking systems. Ordinarily, the majority of these systems support regular flat screens up to 27-inch in size. Even though such paradigms have enabled us to gain considerable knowledge on infant’s visual behaviour in response to experimental stimuli, more ecologically valid investigations in wider visual areas have been limited. Head-mounted eye tracking systems have been an alternative method that allows free head and body movements. However, some developing populations do not respond positively to these systems and they may easily remove the device. Further, these devices are complex to set up with the result that attrition rates are high (Corbetta, Guan & Williams, 2012). In the current study, we aimed to test infants in a wider visual field without any physical constraint for the participant. To do so, we adapt a 4-camera Smart Eye Pro remote system for infant research. We could accurately identify gaze direction in a wide field of view of almost 130° on a 49-inch curved screen area. In addition, we were able to consider both eye and head components which made up the infant’s orienting behaviour. We used this new method to record data from 35 9-month-old participants and we obtained usable data from 26 of them, with an attrition rate of 25.7%. Gaze and head movements were recorded while infants looked at visual targets appearing initially at 60° in their visual periphery before moving towards the centre of the screen. Firstly, in order to track this unusual display, we prepared a world model of our three-dimensional environment (FIG.1). Calibration in such a wide field of view was challenging for infants and we combined an online system calibration with an offline gaze calibration. For the latter, we prepared a simple interface that allows the experimenter to view the recorded data on a frame by frame basis. The experimenter can save multiple frames in which it is safe to assume that the participant is looking at the visual target on screen. The new calibrated gaze coordinates are calculated by averaging all the calibration points and this estimation is displayed on the interface (FIG.2). We obtained a reliable estimation by using 6 offline calibration points per participant. Interestingly, we were also able to track the head position throughout the entire recording and monitor the infants’ eye distance from the display. Although this distance was initially set at 40cm, the variation throughout the recording was high, ranging from 29.2 to 56.5cm (M = 42.2cm, SD = 4.7). Importantly, this paradigm could accommodate this variation and still produce meaningful results. We believe that this approach to tracking young participants’ natural looking behaviour in less constrained environments covering a wider visual area is a promising methodological advance for developmental research.

Authors

• Chiara Capparini, Lancaster University
  Presenting Author

• Michelle To, Lancaster University
  Non-Presenting Author

• Clément Dardenne, Smart Eye AB
  Non-Presenting Author

• Vincent Reid, The University of Waikato
  Non-Presenting Author