Quality of Interaction Arising from Augmented Reality Content: A Comprehensive Study

Current augmented reality (AR) head-mounted dis-plays (HMDs) have rapidly developed with the users' requirement to expand the real-world experience to the virtual world. To bridge the real and virtual space, mid-air hand gestures have been regarded as a de-facto interaction method for AR systems. However, providing natural interaction is still limited and mea-suring the quality of interaction (QoI) has received little attention. In this research, we explore a comprehensive study for perceptive QoI in AR HMDs, focusing on frequently used object interaction tasks. To quantitatively analyze the degree of QoI, we develop an AR-QoI database (e.g., select, translation, and rotation) with 72 content scenes that include mutually independent attributes. A novel protocol for QoI evaluation was designed to collect robust subjective opinions in conjunction with object data from 32 participants. Through a systematic formative study, we identify challenges that the user might face when interacting with un-familiar motion. Moreover, we discover meaningful relationships between various combinations of interaction types and the degree of QoI by clustering scene attributes. Notably, the constructed dataset contains a number of ground-truth labels that correspond to each AR scene. Through rigorous statistics evaluation, we demonstrate that our framework is reasonable for measuring QoI.