KNOWLEDGE-ENHANCED MOBILE VIDEO BROADCASTING (KMV-CAST) FRAMEWORK WITH CLOUD SUPPORT
ABSTRACT
The convergence of mobile communications andcloud computing facilitates the cross-layer network design andcontent-assisted communication. Mobile video broadcasting canbenefit from this trend by utilizing joint source-channel codingand strong information correlation in clouds. In this paper, aknowledge-enhanced mobile video broadcasting (KMV-Cast) isproposed. The KMV-Cast is built on a linear video transmissioninstead of traditional digital video system, and exploitsthe hierarchical Bayesian model to integrate the correlatedinformation into the video reconstruction at the receiver. Thecorrelated information is distilled to obtain its intrinsic features,and the Bayesian estimation algorithm is used to maximize thevideo quality. The KMV-Cast system consists of both likelihoodbroadcasting and prior knowledge broadcasting. The simulationresults show that the proposed KMV-Cast scheme outperformsthe typical linear video transmission scheme called Softcast, andachieves 8dB more of the peak signal-to-noise ratio (PSNR) gainat low-SNR channels (i.e., -10dB), and 5dB more of PSNR gain athigh-SNR channels (i.e., 25dB). Compared to traditional digitalvideo system, the proposed scheme has 7dB more of PSNR gainthan JPEG2000+802.11a scheme at 10dB channel SNR.
PROPOSED SYSTEM:
we propose a brand-new video transmissionframework, which consists of two functions: (1) correlatedinformation extraction, and (2) utilization of such informationfor fast video recovery at the receiver. We first search thecorrelated information from the cloud based on certain criteria. Then we make full use of the correlated information forvideo data reconstruction at the receiver.A conventional scheme, called Softcast, is a jointsource-channel coding scheme. It avoids quantization, entropyencoding and channel encoding. The initial motivation ofSoftcast is to overcome the cliff effect of wireless videotransmission. But it does not take the correlation informationinto consideration. In our previous work, the correlatedinformation assisted video transmission called DaC-RAN, wasproposed in pseudo analog wireless video transmission. Itcan make full use of correlated information to improve thereconstructed image/video quality. However, the peak signalto-noiseratio (PSNR) of the reconstructed images can not beimproved linearly with the increase of channel signal-to-noiseratio (SNR), due to the mutual interference in the reconstructedimages.In a nut shell, the contributions of this paper include thefollowing three aspects:(1) The correlated information in cloud is represented asthe prior knowledge in Bayesian learning for efficientvideo/image reconstruction. A hierarchical Bayesianmodel is proposed to describe the relationship betweenthe transmitted video and the correlated information.(2) We then design the prior knowledge distilling schemeas well as the Bayesian reconstruction algorithm, withthe goal of maximizing the reconstructed video qualityin terms of PSNR. The most correlated informationis extracted as the prior knowledge to facilitate videorecovery.(3) A novel system of knowledge-enhanced mobile videobroadcasting, called KMV-Cast, is proposed. In such asystem the likelihood and prior knowledge are broadcastedseparately.
EXISTING SYSTEM:
In this section, we analyze the drawbacks of non-lineartransform used in traditional digital mobile video systems.Then, we briefly describe the pseudo analog video transmissionarchitecture, which uses a linear transform and addressesthe limitations of traditional digital mobile video systems.The traditional mobile video systems have the source chan-nel separation problem, see Fig.1. Video codecs use entropycoding to remove the data redundancy as much as possible,while error protection codes add back certain redundancy inorder to protect the video from channel interference and packetloss.Source channel separation is not an optimalstrategy for multicast/broadcast applications where mobileusers have different channel qualities. Since source codingturns real-value video signals into bit sequence, the physicallayer has no clue about the video content at all. Thus itis difficult for the physical layer to make good use of thecorrelated image/video contents to assist with video recovery
CONCLUSIONS
In this paper, the knowledge-enhanced mobile video broadcastingframework called KMV-Cast has been proposed toimprove the video quality in multicast applications. The maindifference from other related works is that the correlatedinformation has been represented as the prior knowledgeand exploited in video reconstruction at the receiver. Thehierarchical Bayesian model has been built to model the relationshipbetween the transmitting video and the prior knowl-edge. Furthermore, the prior knowledge distilling methodhas been given to extract the most correlated informationto maximize the PSNR of reconstructed video. Hence, theKMV-Cast framework includes likelihood broadcasting andprior knowledge broadcasting. The simulation results haveshown that the proposed KMV-Cast performs better than otherexisting ones, especially under low SNRs (i.e., -10dB, -5dB,and 0dB). Furthermore, with highly correlated information, thereconstructed video quality improves significantly, i.e., morethan 7dB of PSNR gain than JPEG2000+802.11a scheme. Inthe future work we will focus on data compression of theprior knowledge, and evaluate the system performance underinaccurate/incomplete prior knowledge.
REFERENCES
[1] Cisco, “Cisco Visual Networking Index: Global MobileData Traffic Forecast Update 2014-2019,” Cisco WhitePaper, Feb. 2015.
[2] Wen Hui, Hai-ying Zhao, Chuang Lin, Yang Yang,“Effective Load Balancing for Cloud-based MultimediaSystem,” in Proc. of EMEIT, Harbin, China, Aug. 12-14,2011, pp. 165-168.
[3] www.prweb.com/releases/2014/04/prweb11776359.htm.
[4] Hoang T. Dinh, Chonho Lee, Dusit Niyato, Ping Wang,“A Survey of Mobile Cloud Computing: Architecture,Applications, and Approaches,” Wireless Communicationsand Mobile Computing, vol. 13, no. 18, pp. 15871611,Dec. 2013.
[5] Neha Gupta, Amit Agarwal, “Context Aware MobileCloud Computing: Review,” in Proc. of INDIACom, NewDelhi, India, Mar. 11-13, 2015, pp. 1061-1065.
[6] Jian He, Di Wu, Yupeng Zeng, Xiaojun Hei, YonggangWen, “Toward Optimal Deployment of Cloud-AssistedVideo Distribution Services,” IEEE Transactions on Circuitsand Systems for Video Technology, vol. 23, no. 10,pp. 1717-1728, Oct. 2013.
[7] Jian He, Yonggang Wen, Jianwei Huang, Di Wu, “On theCostCQoE Tradeoff for Cloud-Based Video StreamingUnder Amazon EC2’s Pricing Models,” IEEE Transactionson Circuits and Systems for Video Technology, vol.24, no. 4, pp. 669-680, Apr. 2014.
[8] Xiaofei Wang, Ted “Taekyoung” Kwon, Yanghee Choi,Haiyang Wang, Jiangchuan Liu, “Cloud-Assisted AdaptiveVideo Streaming and Social-Aware Video Prefetchingfor Mobile Users,” IEEE Wireless Communications,vol. 20, no. 3, pp. 72-79, Jun. 2013.
[9] Dipayan Dev, Krishna Lal Baishnab, “A Review andResearch Towards Mobile Cloud Computing,” in Proc.of IEEE MobilCloud, Oxford, United Kingdom, Apr. 811,2014, pp. 252-256.
[10] Wu-Hsiao Hsu, Chi-Hsiang Lo, “QoS/QoE Mappingand Adjustment Model in the Cloud-based MultimediaInfrastructure,” IEEE Systems Journal, vol. 8, no. 1, pp.247-255, Mar. 2014