Deutsch

Quality of Experience

Focus areas:

  • Quality of Experience for speech, video, gaming, and XR
  • Network impairments and their perceptual consequences
  • Subjective test methods and questionnaire design
  • Statistical modelling of perceived quality
  • Standardization at ITU-T, ISO/IEC and DKE

Quality of Experience links what a system does technically to what a person perceives — and therefore to whether they trust it and keep using it. We run controlled subjective tests, model the results statistically, and feed them back into system design. This line carries our standardization work: ITU-T Recommendation P.812 on subjective test methods for interactive virtual reality applications was co-developed here, one of 17 contributions to ITU-T Study Group 12, ISO/IEC JTC 1/SC 29/WG 2 (MPEG) and DKE at DIN and VDE. The methods are the point — a measurement that cannot be reproduced in another laboratory cannot become a standard.

Related projects:

Transfer & Standardization

Measuring what people actually experience

Quality of Experience is the part of a system that cannot be read off a specification. We run controlled subjective tests and crowdsourced studies to find out what a person notices, what annoys them and what they stop noticing after a minute, and we report the conditions under which each result holds. The point of the work is not a score. It is knowing which technical decision changes the experience, by how much, and for whom — the answer that a mean opinion score on its own never gives.

Methods that become standards

A test method is only useful if other laboratories can repeat it. The lab has contributed to ITU-T Study Group 12 over several years, most visibly to Recommendation P.812 on subjective test methods for interactive virtual reality applications, developed from the P.IntVR work item through a series of contributions on test design, quality dimensions and social aspects. This is the slowest part of the work and the part with the longest reach: a method in a Recommendation outlives every study that produced it.

When the network gets in the way

Immersive systems are rarely local. We measure what happens to emotion, presence and game experience when network quality drops in virtual reality cloud gaming, and how impairments in video streaming show up in the face and the gaze of the person watching. The work on estimating quality of experience during video streaming from facial expression and gaze direction received the Best Journal Paper Award of the IEEE Communications Society's Multimedia Communications Technical Committee.

Beyond the questionnaire

Asking people is necessary and not sufficient. We pair subjective ratings with physiological and behavioural measures, including electroencephalography, where the P300 component turned out to indicate context-dependent changes in speech quality that phonological analysis alone does not explain. Where an existing instrument does not fit — too slow, too verbal, too coarse — we build and validate one, and publish it for others to use.

Selected work