MOAVI (Monitoring Of Audiovisual Quality by Key Indicators)
The MOAVI group is an open collaborative for developing No Reference (NR) models for monitoring audio-visual service quality. The goal is to develop a set of key indicators (e.g. blocking effects, blurring effects, freeze/jerkiness effects, ghosting effects, slice video stripe errors, aspect ratio problems, field order problems or photosensitive epilepsy flashing effects, silence, clipping) describing service quality in general and to select subsets for each potential application. Therefore, the MOAVI project concentrates on models based on key indicators contrary to models predicting overall quality.
The MOAVI is a complementary, industry-driven alternative of Quality of Experience (QoE) to measure automatically the audiovisual quality by using perceived simple indicators. The perceived indicators should have a robust prediction performance with a minimum operational restriction. Targeted services include Video on Demand (VoD), live broadcast services (Satellite, IPTV, Digital Terrestrial Television). The audio/video indicators could be based on analyzing the video signal only, or by using parametric (bitstream) or hybrid measurements (bitstream + video signal).
Measurements could be applicable according to the availability of access points along the video chain (the video head-end, server of content delivery, terminal, encrypted or not).
The MOAVI activities are split into four steps:
- Maintain a list of potential real-world applications for audio-visual quality monitoring.
As a result, some additional artifact definitions could be submitted to ITU-T G.100.
- Identify the main audio and video indicators taken into account in the customer acceptability.
The contributors are invited to suggest the most representative perceived indicators. During this step, participants can also propose some appropriated subjective tests for each indicator in relationship with user acceptability of ITU-T. G100.
- Design the indicators according to 3 categories.
According to the result obtained at the previous step, participants help to design each indicator for one or more of the categories:
a. Based on the audio-visual signal
b. Based on the parametric
c. Based on hybrid
If possible, the models will be designed by using video sequences collected in operational conditions.
- Performance evaluation of the indicators.
The performance of each indicator will based on the aim of maximization of the true prediction (true positives/negatives) and minimization of the false prediction (false positives/negatives). The statistical instruments may include: Precision, Recall, Specificity, Sensitivity, Accuracy, F1-score, etc. Furthermore, if DCR-like scores are collected, regular MOS-like statistical analysis will be applicable as well.
For more information, see the following two papers:
Also, see the following webpage:
AGH University of Science and Technology Video Quality Resources (video quality research, metrics, paper references).
The subpage with MOAVI indicators is: http://vq.kt.agh.edu.pl/metrics.html.