scope:

This Recommendation describes a derivation procedure for contribution values that provides information about a relationship between a quality-influencing factor (i.e., media quality levels and stalling) and media session quality of adaptive audiovisual streaming services.

In streaming services, the quality experienced by the end user can be assessed by using quality estimation models such as:

- A metadata-based model, which uses audio and video bitrates, framerate, resolution ITU-T P.1203.1 mode 0, [ITU-T P.1203.2], [ITU-T P.1203.3];

- A bitstream-based model that uses meta-data and any information from the video bitstream ITU-T P.1203.1 mode 3, [ITU-T P.1204.3], [ITU-T P.1203.2], [ITU-T P.1203.3];

- A pixel-based model that uses information derived from reference video [ITU-T P.1204.4], [ITU-T P.1203.2], [ITU-T P.1203.3];

- A hybrid metadata-based and pixel-based model [ITU-T P.1204.5], [ITU-T P.1203.2], [ITU-T P.1203.3].

These models estimate a mean opinion score (MOS) on a five-point absolute category rating scale [ITU-T P.910] as a global multi-media MOS score [ITU-T P.911] by taking into account qualityinfluencing factors such as media quality level (i.e., set of audio-coding or video degradations due to coding, spatial re-scaling or variations in video frame rates) and stalling (i.e., delivery degradations due to initial loading delay and rebuffering at the client) [ITU-T P.1203]. In this Recommendation, the output of these quality estimation models (i.e., global multi-media MOS score) is referred to as the final media session quality score as specified in [ITU-T P.1203].

The derivation procedure described provides contribution values that show how much each media quality level and stalling affects (i.e., decreases) the final media session quality score provided by quality estimation models such as [ITU-T P.1203] as previously described for each media session. This contribution value is calculated as the degree of quality degradation due to each media quality level and stalling by comparing with the maximum final media session quality score (i.e., the highest audiovisual quality and no stalling during the media session) based on Shapley theory, which is a key solution concept in cooperative game theory for distributing profits in a game among participants [b-Shapley]. Here, for example, different media quality levels selected or stalling occurrences during the media session result in different contribution values of each media quality level and stalling. Also, if the final media session quality score of a certain media session is different because of the use of a different quality estimation model, the contribution values of each media quality level and stalling are also different.

The derivation procedure described is agnostic to the specified quality estimation models such as the ITU-T P.1203 type of model. This means that, although a metadata-based model (i.e., ITU-T P.1203 mode 0), bitstream-based model (i.e., ITU-T P.1203 mode 3), pixel-based model (i.e., [ITU-T P.1204.4]) and hybrid model (i.e., [ITU-T P.1204.5]), and their respective appendices for long term integration, generally exist to estimate the final media session quality score, this Recommendation can be used with all types of quality estimation model. As examples, the procedures when ITU-T P.1203 mode 0 (metadata-based model), ITU-T P.1203 mode 3 (bitstream-based model) and [ITU-T P.1204.4] (pixel-based model) are used as quality estimation model are described in Appendices I, II and III, respectively.

This Recommendation primarily applies to the analysis of the quality aspect of progressive download or adaptive audiovisual streaming so that operators can ascertain degradation factors by deploying the derivation procedure described in end-point locations, mid-network monitoring points or a streaming server.