Mukundan Venkataraman

Inferring video QoE in real time

Inferring the subjective perception of a video stream in real time continues to be a stiff problem. This article presents MintMOS: a lightweight, no-reference, loadable kernel module to infer the QoE of a video stream in transit and offer suggestions to improve it. MintMOS revolves around one-time offline construction of a k-dimensional space, which we call the QoE space. A QoE space is a known characterization of subjective perception for any k parameters (network dependent/ independent) that affect it. We create N partitions of the QoE space by generating N video samples for various values of the k parameters and conducting subjective surveys using them. Every partition then has an expected QoE associated with it. Instantaneous parameters of a real-time video stream are compared to the precomputed QoE space to both infer and offer suggestions to improve QoE. Inferring QoE is a lightweight algorithm that runs in linear time. We implemented MintMOS by creating an actual QoE space using three parameters and 27 partitions by conducting surveys with 77 human subjects. In a second set of surveys using 13 video clips, MintMOSs predictions were compared to 49 human responses. Results show that our MOS predictions are in close agreement with subjective perceptions.

Evaluating Quality of Experience for Streaming Video in Real Time

We present a scalable, lightweight, no-reference framework to infer video QoE. Our framework revolves around a one time offline construction of a k-dimensional space, which we call the QoE space. The k-dimensions accommodate k parameters (network-dependent/independent) that potentially affect video quality. The k-dimensional space is partitioned to N representative zones, each with a QoE index. Instantaneous parameter values are matched with the indices to infer QoE. To validate our framework, we construct a 3-dimensional QoE space with bit-rate, loss, and delay as the principal components. We create 18 video samples with unique combinations of the 3 parameters. 77 human subjects rated these video samples on a scale of 1 to 5 to create the QoE space. In a second set of survey, our predicted MOS was compared to 49 human responses. Results show that our MOS predictions are in close agreement with subjective perceptions. An implementation of our framework on standard Linux PC shows we can compute 20 MOS calculations per second with 3 parameters and 18 partitions of the QoE space.

Quantifying video-QoE degradations of internet links

With the proliferation of multimedia content on the Internet, there is an increasing demand for video streams with high perceptual quality. The capability of present-day Internet links in delivering high-perceptual-quality streaming services, however, is not completely understood. Link-level degradations caused by intradomain routing policies and inter-ISP peering policies are hard to obtain, as Internet service providers often consider such information proprietary. Understanding link-level degradations will enable us in designing future protocols, policies, and architectures to meet the rising multimedia demands. This paper presents a trace-driven study to understand quality-of-experience (QoE) capabilities of present-day Internet links using 51 diverse ISPs with a major presence in the US, Europe, and Asia-Pacific. We study their links from 38 vantage points in the Internet using both passive tracing and active probing for six days. We provide the first measurements of link-level degradations and case studies of intra-ISP and inter-ISP peering links from a multimedia standpoint. Our study offers surprising insights into intradomain traffic engineering, peering link loading, BGP, and the inefficiencies of using autonomous system (AS)-path lengths as a routing metric. Though our results indicate that Internet routing policies are not optimized for delivering high-perceptual-quality streaming services, we argue that alternative strategies such as overlay networks can help meet QoE demands over the Internet. Streaming services apart, our Internet measurement results can be used as an input to a variety of research problems.

Effects of internet path selection on video-QoE: analysis and improvements

This paper presents large-scale Internet measurements to understand and improve the effects of Internet path selection on perceived video quality, or quality of experience (QoE). We systematically study a large number of Internet paths between popular video destinations and clients to create an empirical understanding of location, persistence, and recurrence of failures. These failures are mapped to perceived video quality by reconstructing video clips and conducting surveys. We then investigate ways to recover from QoE degradation by choosing one-hop detour paths that preserve application-specific policies. We seek simple, scalable path selection strategies without the need for background path monitoring. Using five different measurement overlays spread across the globe, we show that a source can recover from over 75% of the degradations by attempting to restore QoE with any

Dynamic Routing Framework for Wireless Sensor Networks

k

Effects of internet path selection on video-QoE

randomly chosen nodes in an overlay, where

Measuring and quantifying the silent majority on the Internet

k

A dynamic reconfigurable routing framework for wireless sensor networks

is bounded by

Traffic Based Dynamic Routing for Wireless Sensor Networks

O(\ln(N))

Spatio-Temporal Analysis of Passive Consumption in Internet Media

. We argue that our results are robust across datasets. Finally, we design and implement a prototype packet forwarding module called source initiated frame restoration (SIFR). We deployed SIFR on PlanetLab nodes and compared the performance of SIFR to the default Internet routing. We show that SIFR outperforms IP-path selection by providing higher on-screen perceptual quality.