Network-based Video Freeze Detection/Prediction in HTTP Adaptive Streaming

With the popularity of HTTP adaptive streaming technology for media delivery technology over mobile and fixed networks, the clients Quality of Experience for HAS video sessions is particularly of interests in network providers and Content Delivery Network providers. But typically, network providers are not able to directly obtain QoE relevant metrics, such as freeze, initial loading time, quality switches, etc, from the client. This paper proposes a scalable machine learning (ML) based scheme that detects/predicts video freezes using a few features extracted from the network-based monitoring data, i.e., a sequence of HTTP GET requests sent from the client. We examine the discriminative features with multi-scale windows for representing the freeze using the criterion of information gain. Four classical classifiers are investigated and the C4.5 decision tree is distinguished because of its simplicity, scalability, accuracy, and interpretability. Our approach for session based offline freeze detection is evaluated on the Apple HTTP Live Streaming video sessions obtained from a number of operational CDN nodes and on the traces of Microsoft Smooth Streaming video sessions acquired in a controlled lab environment. Experimental results show that an accuracy of about 97% can be obtained for the detection of the existence of a video freeze, a long video freeze and multiple video freezes. The segment based online freeze prediction is examined on the traces collected from the OpenFlow settings. Results suggest that more than 30% of freezes can be foreseen one segment in advance.