Towards better video services: An EEG-based interpretable model for functional quality of experience evaluation

Since emerging video services can provide emotional and social value to users, the setting of their functional parameters directly affects human cognitive and affective states, further influencing video services’ quality of experience (QoE), which we call functional QoE (fQoE). FQoE is highly dependent on human subjective perceptions and the reasons for its generation are important for service providers to optimize video services. However, existing fQoE research methods are unable to perform quantitative assessment and lack interpretability. Electroencephalogram (EEG) signals have the advantage of being difficult to disguise, and contain rich brain activity information, gaining more attention from researchers nowadays. Based on EEG, we propose an interpretable model to evaluate fQoE, and the model is tested on a self-built dataset for bullet chatting video (BCV) service. Our model can effectively fuse single electrode and multi-electrode features from EEG, and introduces Graph-based Brain-area Perception Network (GBPN) for extracting fQoE sensitive brain areas, achieving satisfactory results. We find brain areas associated with fQoE caused by different functional parameters of BCV. To sum up, our fQoE model enables quantitative assessment with neurophysiological interpretability of fQoE, providing a scientific basis for the optimization and development of video services.