CasualGaze: Towards Modeling and Recognizing Casual Gaze Behavior for Efficient Gaze-based Object Selection

TL;DR

CasualGaze is presented, a novel eye-gaze-based target selection technique to support natural and casual eye-gaze input, employing a bivariate Gaussian distribution model along with temporal compensation and voting algorithms for robust target prediction.

Abstract

We present CasualGaze, a novel eye-gaze-based target selection technique to support natural and casual eye-gaze input. Unlike existing solutions that require users to keep the eye-gaze center on the target actively, CasualGaze allows users to glance at the target object to complete the selection simply. To understand casual gaze behavior, we studied the spatial distribution of casual gaze for different layouts and user behavior in a simulated real-world environment. Results revealed the impacts of object parameters, the speed and randomness features of casual gaze, and special gaze behavior patterns in "blurred areas". Based on the results, we devised CasualGaze algorithms, employing a bivariate Gaussian distribution model along with temporal compensation and voting algorithms for robust target prediction. Usability evaluation study showed significant improvements in recognition and selection speed for CasualGaze compared with two baseline techniques. Subjective ratings and comments further supported the preference for CasualGaze regarding efficiency, accuracy, and stability.

Figures (18)

• Figure 1: The target location set for Study 1 (left). Three heights and eight horizontal angles are set with the user as the center. The red cone indicates the target direction in a single trail. The offset of the user's gaze relative to the target direction(right). Calculate the horizontal ($\phi$) and vertical ($\theta$) angles of the user's gaze (red line) and the target direction (blue line). The difference between these two angles represents the offset of the user's gaze relative to the target.
• Figure 2: Offsets at different heights (left). Black dots are target centers and red dots are offsets; Standard deviation of the distribution in different horizontal directions (right). The length of the line segment in different directions represents the value of the standard deviation.
• Figure 3: Task progress of Study 2. After the three-second countdown, two spheres appear in front of the user (left). The relative size, location, and distance of the two spheres are randomly generated. After confirming the location, the user pulls the trigger, a red cone appears and indicates to select the target (middle). The user select the target with gaze and pulls the trigger again. The selected target lights up to give the user feedback (right).
• Figure 4: The distribution under different relative locations. Different colors represent the Gaussian distribution under different relative locations. The black point represents the center of the Gaussian distribution after the shift. It shows that the relative location causes a shift in the center of the Gaussian distribution, and the shift levels in the two directions are different.
• Figure 5: Gaussian distribution of different goal target sizes. The red circles represent the target itself, the red dots represent the target edges, and the black dots represent the radius of the Gaussian distribution.