Practical Channel-adaptive Video Streaming with Fountain Codes

Abstract

Abstract-Video streaming is sensitive to packet loss, which can severely damage the quality of the received video. Video communication systems that rely on application-layer forward error correction (FEC) to combat packet loss are particularly suitable for pervasive computing because they can be used on top of any existing network architecture. However, since in heterogeneous environments network conditions are unpredictable, determining the right amount of redundancy introduced by the channel encoder is not obvious. This paper presents a practical implementation of a unicast video streaming system that solves this problem by using a rateless code and receiver feedback. In real simulations over the Internet our solution outperformed a standard approach based on fixed-rate forward error correction. For an Internet connection Konstanz-BeijingKonstanz and the standard Foreman sequence compressed with the H.264 video coder, the gain in average peak signal to noise ratio exceeded 3.5 decibels at 90 kilobits per second.

State of the Art on Video Streaming and Applications

“…21), at 89.86 kbps, whereas scheme (a) achieved an average PSNR of 32.02 dB, scheme (c) reached an average PSNR of 28.36 dB at 90.04 kbps (Ahmad et al, 2008a). Furthermore, scheme (c) achieved better PSNR than (d) and this is because the packet loss rate is rapidly changing, so as a result the packet loss rate prediction from only the previous source block interval (not from the histogram) will be not efficient (Ahmad et al, 2008a).…”

“…Furthermore in (Ahmad et al, 2008a), they implemented (Ahmad et al, 2007) and tested the performance of the live unicast video streaming system under H.264/AVC video compression standard. Then they make a comparison (based on Peak Signal to Noise Ratio (PSNR)) between the following schemes: a) Algorithm 1: Proposed in (Ahmad et al, 2007) and based on prior packet loss rate histogram b) Algorithm 2: Practical version of Algorithm 1 (Ahmad et al, 2008a), also the packet loss rate histogram is computed in real-time c) Static: Fixed rate code until acknowledgement come from the receiver d) Adaptive: Here the code rate is selected depending on the lost packets rate that observed in the previous source block transmission interval only (not based on the packet loss rate histogram) e) Without FEC: This scheme did not use FEC Actually, Fig. 21 shows that schemes (a) and (b) achieved better PSNR than (c),(d) and (e).However, scheme (b) had a slightly worse performance than scheme (a) and this because it did not rely on prior observations (or histogram) of the packets loss rates (Ahmad et al, 2008a).…”

“…In addition, the transmission strategies and packet loss rate histogram (as shown in Fig. 18) in (Ahmad et al, 2007;2008a) lead to decreasing the FEC coding bandwidth overhead while keeping good reconstruction of the video streaming and hence results in enhancing the average PSNR and decreasing the congestion level. Ahmad et al (2010) they did a lot of experiments to show that the performance of solutions in (Ahmad et al, 2007;2008a) is close to optimal compared to other FEC solutions for live or prerecorded video streaming over packet erasure channel.…”

Adaptive Unicast Video Streaming With Rateless Codes and Feedback

“…The optimization does not rely on only one packet loss rate but on a histogram of previously observed packet loss rates. This paper synthesizes and extends preliminary results presented in [9] and [10]. Specifically, we provide an in-depth discussion of related work, give a detailed description of the proposed transmission strategy and the algorithm to optimize it, and significantly expand the experimental work.…”

Media Coding for Streaming in Networks with Source and Path Diversity

“…However, [23] makes several unrealistic assumptions such as the employed codes have a fixed overhead, the pdf of the channel loss rate is known and the communication is performed through a reliable channel by a single symbol message. These shortcomings are addressed in [24] where the channel conditions and thus the pdf are derived from real time measurements. Raptor codes are replaced by LT codes for utilizing the low cost belief propagation decoder.…”