Holo-QoI: A Human Factor-Based Dataset and Prediction Framework for Assessing Quality of Interaction in Augmented Reality

Seongjean Kim, Seonghwa Choi, Sanghoon Lee

Abstract: With the revolution of Augmented Reality (AR) Head Mounted Display (HMD), AR technology enables the provision of immersive virtual content using various interaction methods. To bridge the real and virtual world, hand-based interactions have been utilized for their simplicity and effectiveness in interacting with AR applications. Thus, assessing the Quality of Interaction (QoI) is vital for enhancing the experience of AR applications, but it has been hindered by the absence of a comprehensive database that includes human factors. In this paper, we construct an interaction dataset, Holo-QoI, which includes 49 sets of human factor metadata with associated human subjective scores. To analyze the contribution of various factors to QoI, we process HMD metadata into human factor-based features (e.g. head, gaze, and hand motion). We then design a QoI prediction framework based on a Transformer architecture to model the relationship between human factor-based features and deep features, which is used to predict the QoI score of an interaction. Through rigorous experiments, we demonstrate that our proposed model achieves an 86% accuracy using the constructed dataset.