Estimating the effect of context on QoE of audiovisual services: Laboratory vs. public places

Multimedia service providers and operators need tools that can estimate the quality of the services delivered to their end-customers. To this end, a variety of quality prediction models has been developed. The models are typically built and trained based on data acquired from user studies conducted in laboratories with precisely defined conditions. While laboratory originated data is accurate and reliable, the performance of Quality of Experience (QoE) models could be possibly further improved by taking into account the real context of use. Working towards that goal, in this paper, we compare results of a laboratory audiovisual quality subjective study and two smaller scale user studies conducted in public place. We also discuss our approach of enhancing the prediction accuracy with context-specific bias functions.