Intelligent Attention-Based QoE Enhancement for VR Interaction: Keyframe Extraction and Resource Allocation

The rapid expansion of multi-user virtual reality (VR) applications, such as VR gaming and the Metaverse, has heightened the demand for bandwidth-efficient, immersive experiences. This paper introduces a customized Quality of Experience (QoE) framework tailored to VR interactions in a sub-6 GHz communication environment, integrating attention-based interaction, keyframe extraction, and principles from the Weber-Fechner Law. The QoE maximization problem is formulated as a Mixed Integer Programming (MIP) problem that jointly optimizes keyframe extraction ratio and resource allocation (i.e., bandwidth and CPU frequency), incorporating fairness. We employ the Deep Deterministic Policy Gradient (DDPG) algorithm as a reinforcement learning strategy. Evaluation with the Motion Capture Database demonstrates that our framework significantly reduces interactive latency, enhances QoE, and maintains fairness, achieving superior performance compared to baseline methods.