Deep Learning Empowered Real-Time Haptic Communications for IoT

The advancement of the Internet of Things (IoT) has enabled a variety of innovative consumer electronics applications. In particular, multi-modal applications integrating audio, video, and haptic interactions have emerged, offering users more immersive experiences and enhanced remote interaction capabilities. However, the distributed and fragmented nature of IoT network resources presents significant challenges for ensuring haptic communications under strict constraints. To address these challenges, this paper proposes a joint coding and network resource scheduling solution for haptic communications in IoT environments. First, we design a scalable and flexible haptic coding scheme that supports efficient and seamless haptic streaming across heterogeneous IoT networks. Second, we introduce a deep learning-based imperceptible Quality of Experience (QoE) recognition method, which uses emotional immersion as an evaluation metric to achieve unobtrusive QoE monitoring. Finally, based on user QoE feedback, we develop an on-demand cooperative network resource scheduling strategy to ensure high QoE, reduce network resource wastage, and optimize the utilization of fragmented IoT resources. Simulation results validate the effectiveness of the proposed architecture and strategies.