ROI-Based Video Transmission in Heterogeneous Wireless Networks With Multi-Homed Terminals

We consider the problem of delivering region of interest (ROI)-coded mobile video streams using limited radio resources. Under the conditions of limited bandwidth and time-varying channel status, the goal is to optimize the transmission latency, while ensuring the quality of the ROI parts. Multi-homing support enables the terminals to establish multiple connections for transmission performance improvement. In this paper, we propose a novel framework for ROI-based video transmission in heterogeneous wireless networks with multi-homed terminals. The framework contains the modules of ROI detector and frame splitter, where macroblocks are categorized based on ROI detection and encapsulated into transforming units. It also includes a channel monitor that keeps track of the status of each communication path and sends feedback signals to the streaming controller for packet-scheduling control; a deep learning method is proposed for channel status prediction. To address the delivery problem, we propose a scheduling approach based on the formulated network model and the rate-distortion model. The scheduling method makes a tradeoff between the transmission delay and the distortion. It also guarantees that packets with ROI content are delivered on paths with sufficient bandwidths and low loss rates. Through comparisons with other scheduling methods, we find that the proposed scheme outperforms the other scheduling methods in terms of improving the quality (peak signal-to-noise ratio), balancing the end-to-end delay, and maintaining the playback fluency.