Quality of Experience Inference for Video Services in Home WiFi Networks

Abstract

Customer experience is becoming of utmost importance for operators in home network management. Hence, a notion of the customers' QoE is vital, but has mostly been neglected in favor of QoS. In this article, we aim to close a significant gap between QoS and QoE in home networks by proposing a framework for inferring QoE from remotely collected network QoS metrics. We focus on video services (e.g., YouTube application) as the main contributor and generator of indoor network traffic. A case study is performed where an experimentally obtained dataset comprising network and application QoS parameters is obtained under varying conditions (i.e., poor coverage, network overload, contention, and interference). Predictive modeling is then used to build a predictor for multiple QoE classes given the network QoS metrics remotely accessible from access points based on industry adopted standards (i.e., TR-181). This enables operators to infer specific QoE metrics using remotely collected passive network measurement with no knowledge of application-specific parameters. We show that the proposed framework achieves accuracy in the range of 85 to 95 percent depending on the QoE class, hence demonstrating the effectiveness and potential of our approach.