Affective-based modelling approaches for quality of experience-based management systems

The ubiquity of multimedia content consumption in human digital lives has made Quality of Experience (QoE) a pivotal factor in user satisfaction and engagement. However, to build personalised QoE models, extensive, costly, and time-consuming subjective studies are required, which are affected by several limitations (e.g., the need to ask for explicit feedback from the user) and are unsuitable for real-time management systems. Thus, novel QoE models based on the analysis of the affective state of the user while consuming multimedia content have gathered particular attention in recent years. Machine learning techniques are essential in this regard for supporting the analysis of vast amounts of user emotional data (e.g., features extracted from facial expressions, speech, and heart rate) and indicating a possible correlation between these features and the perceived QoE. In this Chapter, affective-based modelling approaches for QoE estimation are discussed. In particular, two QoE estimation systems based on facial-related features are presented, which are developed for video streaming and WebRTC-based applications, respectively. The QoE estimation performance achieved by these models demonstrates the potential of affective-related indicators in estimating the QoE for multimedia services.