SVC-based and Flow Assignment Strategy for Multi-path Concurrent Video Transmission with Low-Latency

The most conventional approaches for multi-path concurrent video transmission with H264/AVC coding are considering the distribution strategy to improve the video transmission quality under strict delay constraints. However, guaranteeing the smooth playback of real-time video is a challenge for conventional solutions due to the fluctuating status and scarce resources of wireless networks and subject to stringent end-to-end latency and packet loss. In order to address these critical issues, this paper leverages the properties of Scalable Video Coding (SVC) coding and joints flow assignment strategies to improve video transmission performance. It effectively aggregates the channel resources in heterogeneous networks and determines different SVC layer streams delivered upon different communication path statuses. In addition, the SVC base layer resources reservation scheme is proposed to reserve sufficient resources for subsequent SVC base layer streams as the current SVC streams are allocated. The semi-physical experiment results show that the proposed algorithm achieves 4–8 dB higher PSNR transmission quality than the conventional schemes with the influence of network time-varying characteristics.

I. Introduction

With the rapid rise of mobile real-time video services, society's demand for video transmission applications is getting higher and higher. For instance, the video transmission applications commonly take an average bandwidth of 6 to 8 Mbps, which exceeds the capacity of most wireless communication systems [1] once the H.264/AVC compressed 1080P video is transmitted. This single network access method will bring a heavy burden for low-latency and high-quality video transmission due to factors such as low-throughput, high-latency, and limited-coverage. The significance of a heterogeneous network is that it could support applications with higher throughput, lower latency, and better error recovery capabilities through the integration of multiple simultaneous access network technologies (such as LTE, WIFI, 5G, etc.) [2]. In recent years, an increasing number of researchers have concentrated their effort on real-time video Concurrent Multipath Transmission (CMT) [3] [4] in heterogeneous wireless networks. Wu et al. [5] proposed a Sub-Frame Level (SFL) algorithm that splits the video frame by modeling the end-to-end delay and distributes it to each path with the water-filling algorithm, which effectively reduces the end-to-end delay of the video frame. Wu and Yuen [6] show an Adaptive Flow Assignment and Packet Scheduling (AFAPS) framework. Considering that data loss will seriously affect video quality, Wu et al. [7] present a quAlity-Driven Multipath TCP (ADMIT) scheme by allying TCP and Forward Error Correction (FEC). This method models the video transmission quality of each communication path and selects the most reliable access network for joint FEC encoding to minimize end-to-end video distortion. Both [6] and [7] suggest the pathChirp algorithm [8]to estimate the available bandwidth of each communication path. The algorithm requires additional probe messages to estimate the bandwidth, which will lead to video transmission efficiency being significantly reduced in the case of network deterioration.