Multidimensional QoE Prediction of WebRTC Video Communication with Machine Learning

Abstract

Web Real-Time Communication (WebRTC) technology has quickly become popular as a video conferencing platform since it enables cost-effective peer-to-peer communication via browsers. In order to test cost-effectiveness of using WebRTC technology, it is necessary to dispose of user feedback through Quality of Experience (QoE) scores, considering various Influence Factors (IFs). Several studies have analyzed the impact of Context IFs (CIFs) on perceived QoE so far. This paper considers the impact of two CIFs, i.e., conversational task and different types of WebRTC-based applications on the QoE of video communication. In this sense, it focuses on several QoE dimensions, i.e., the overall satisfaction, efficiency, ease of use, and acceptance of WebRTC. QoE scores were collected by survey, and then statistically analyzed using Analysis of Variance (ANOVA) and Pearson correlation. Multi-Layer Perceptron (MLP) Artificial Neural Network (ANN) model has been developed in order to predict QoE based on its dimensions (i.e., overall satisfaction, efficiency, ease of use, and acceptance). Results indicate that the developed MLP model has a medium strong ability to make accurate QoE predictions.