Deep Learning-Aided QoE Prediction for Virtual Reality Applications Over Open Radio Access Networks

Abstract

Nowadays, innovative applications in the field of virtual reality (VR) are being developed, attracting the interest of both academia and industry. Wireless VR applications focus on various aspects of daily life, such as smart education, entertainment, healthcare, tourism, architecture, automotive, and industrial automation. All these inherently interactive applications that aim to create immersive experiences for users are closely related to the concept of quality of experience (QoE), which expresses the quality of a service as perceived by end-users. In this paper, we develop an objective QoE prediction model based on deep learning techniques. The prediction model examines the impact of wireless network operation on the quality of VR 360-degree video streaming. It is based on an encoder-decoder long short-term memory (LSTM) neural network and is able to predict in real-time the overall transmission-related QoE value using only measurable quality of service (QoS) parameters. The prediction model is tested and evaluated on an open radio access network testbed with interfaces based on the O-RAN specifications.

Keywords

• Deep learning
• deep neural network (DNN)
• long short-term memory (LSTM)
• open radio access network (Open RAN)
• quality of experience (QoE)
• QoE prediction