Quality of Multimedia Experience Prediction using Peripheral Physiological Signals

This paper proposes the utilization of physiological signals for quality of experience (QoE) assessment by employing machine learning and deep learning classifiers. Accurately predicting user QoE by analysing physiological signals holds significant potential in diverse fields, including human-computer interaction, healthcare, and education. To predict various QoE factors from physiological signals, the experiments were conducted on two datasets: SoPMD Dataset 1 and SoPMD Dataset 2. The bidirectional long-short-term memory (BLSTM), support vector machine, k-nearest neighbour and random forest algorithms were evaluated using fused electrocardiogram and respiration signals to predict subjective QoE scores, including perceived quality levels, user preference, and the sense of presence. The results demonstrate the effectiveness of the models, with BLSTM emerging as the top-performing algorithm across most experiments, achieving high classification F1-scores. These findings suggest that the physiological signals can be effectively used in the classification of subjective QoE scores.