TECHNICAL
video and Spirent’s Video Media Server that simulates telemetry data while also streaming a gaming video to the mobile device. Spirent improves on the ITU method and replaces estimation of video quality from video parameters (parametric) with actual measured video quality of pixels (non-reference). A Methodology You Can Trust Achieving a machine-driven video quality assessment solution allows developers and service operators to objectively and repeatedly evaluate their ability to deliver high quality streaming buffered and real- time video. With a proven correlation to the state-of-the-art full-reference VMAF solution, the Spirent Umetrix Video Non- Reference Compression model opens the door for content services and device OEMs to measure how well their network or device delivers content from OTT streaming services such as Netflix and Hulu. They can benchmark their progress over time and compare themselves to their competition. “Stare and compare” can be replaced by a robust and scalable solution that works in the field, in drive testing, and in QA and development labs.
Chart 2 – A Pearson correlation of over 90% between VMAF and Umetrix is achieved, giving high confidence that the Umetrix Video Non-Reference Compression model is producing excellent video quality scores.
Can this apply to mobile gaming?
in the baseline data set. Using a Pearson correlation, we would hope to achieve a correlation of greater than 80% across the entire baseline data set. That would give us confidence that our non-reference score offers a very good method of scoring video content without employing a reference for comparison. In fact, due to the attributes of the underlying BRISQUE machine-learning algorithm and the three terabytes of images in the training data set created by Spirent, we have achieved a Pearson correlation of over 90% between VMAF and Umetrix. This gives us high confidence that the Umetrix Video Non-Reference Compression model is producing excellent video quality scores across a very wide variety of scene types and compression levels.
Mobile gaming is a fast-growing segment of the overall gaming industry, so how can we apply Spirent NR algorithms to help test QoE in these environments? Spirent’s process uses an emulated approach to measure the mobile gaming experience and delivers a video QoE score based on dual factors, not just IP packets. Key KPI components are measured, including uplink and downlink telemetry latency and loss, as well as video frame freezes and MOS. The ITU-T G.1072 (encoding parameters, network parameters, and game classification) is used to calculate the user’s QoE using video quality metrics and telemetry metrics based on a derived video captured directly from streaming
www.spirent.com
Volume 47 No.4 DECEMBER 2025
95
Made with FlippingBook - Online magazine maker