The Xperium QoE models are at the core of Avvasi solutions and can be deployed at any point in the delivery chain, at the server, in the network or on the client. This unique technology offers scalable and accurate, passive measurement and active management capabilities based on video QoE for streaming video content such as Netflix®, YouTube™ and Hulu™. This innovative technology provides powerful analytics, assurance, management and service creation capabilities for service providers.
For voice calls, MOS scores (shown below) and associated models are widely accepted quality measures. Xperium QoE models offer a similar quality measure for streaming video, one that accurately reflects the subscriber’s perceived experience with the video content.
Xperium QoE models are based on psycho-visual models, using best practices from the field of vision science, video quality measurement and modeling. These models are developed following many subjective studies around video QoE, which were conducted both independently and with third party research labs and standards bodies around the world.
Video QoE Factors
Video QoE is a perceptual measure which should reflect a subscriber’s satisfaction with their video streaming experience. A QoE metric needs to take into consideration all of the impairments that affect the video quality in the service delivery chain. For streaming, there are two main classes of impairments: delivery issues including delays, and visual quality issues including encoding loss and display impact. While a single QoE score is valuable, it quickly becomes important to understand if a poor experience is due to delivery issues or visual quality issues or both.
Xperium QoE takes a divide-and-conquer approach, treating each class independently. Xperium QoE models use the served content quality, the effects of network degradations, and device screen size and capabilities to produce a Delivery, Presentation, and overall Session QoE score that accurately reflect the real subscriber experience.
XperiumTM Delivery QoE
The vast majority of today’s streaming video services are delivered reliably via HTTP over TCP, where the noticeable delivery-related artifacts are limited to startup delay, stalling and, for adaptive protocols, switching to low bitrate streams (to reduce stalling). The problem of estimating a Delivery QoE can therefore be reduced to predicting the perceptual impact of these delivery-related impairments.
The Xperium™ Delivery QoE model is a parameterized behavioral model that takes these impairments into account and provides a score which accurately predicts a viewer’s QoE at any point during a clip.
XperiumTM Presentation QoE
Ignoring delivery-related issues, the sampling, display and encoding processes are the most important factors contributing to QoE for streaming video. The problem of estimating a Presentation QoE can therefore be reduced to predicting the perceptual impact of sampling, encoding and display-related impairments.
The Xperium™ Presentation QoE model measures the quality level of a media session, taking these impairments into account.
XperiumTM Session QoE
Session QoE builds upon the Delivery QoE and Presentation QoE scores to present, in a single score, an overall QoE score for each media session. This is important for service providers who are interested in understanding the customer experience as a single score, and for correlating to parameters measured by external systems.
Advantages of Xperium QoE
Built upon unique video encoding quality models, human behavioral models and device viewing models, Xperium QoE models were developed using best practices from the quality assessment field of image and video science. They are the only models developed from and validated by subjective studies, both independently and through third party labs such as the University of Texas Austin (UTA) Live Lab.
Xperium™ QoE models are from the class of bitstream-layer no-reference methods. This enables detailed content analysis at affordable computational complexity. This type of model is significantly less complex than competing full-reference methods or pixel-layer no-reference methods, which require baseband video processing, while achieving similar levels of accuracy.
Xperium™ QoE models achieve industry-leading correlation with human subjective scores. These models are significantly more accurate than competing packet-layer no-reference methods, which cannot account for scene or content complexity.