Cloud-assisted solution for crowdsourced live streaming systems

个人和移动计算的进步使人们通过各种社交媒体网络服务相互联系，并在这些网络上分享图像和视频变得越来越方便。特别是，通过移动设备进行众包直播已经成为一种流行的应用，任何用户都可以通过网络直播视频，其他用户可以观看这些视频并与内容提供商或其他观众互动。这些电视网中的一些受欢迎的节目很容易吸引数千甚至数百万的观众。这给众包直播服务提供商带来了几个重大挑战。首先，众包贡献者是地理分布的。其次，众包商往往比专门的内容提供商更具活力。第三，由于用户设备和网络条件的异构性，最终用户的服务质量(Qos)很难得到保证。在这个项目中，我们将与我们的行业合作伙伴合作，充分利用云计算和视频编码技术，开发一种针对众包直播系统的云辅助解决方案。首先，我们将为众包直播设计一个通用的高性价比框架，以支持大规模分布的视频众包商。其次，考虑到广播者和观众的异构性，为系统设计了转码和码率自适应策略。第三，将采用速率控制和播放缓存优化来进一步提高终端用户的服务质量。最后，我们将探索一种混合云和对等系统，作为提高众包直播性能的可能方法。

The advances in personal and mobile computing make it increasingly convenient for people to be connectedwith each other via various social media network services, and share images and videos over these networks. Inparticular, crowdsourced live streaming via mobile devices has become a popular application, where any usercan broadcast live videos through the network, and other users can watch these videos and interact with thecontent providers or other viewers. Some popular events in these networks could easily attract thousands oreven millions of viewers. This has posed several major challenges to the service providers of crowdsourcedlive streaming. First, crowdsourced contributors are geo-distributed. Second, the crowdsourcers are often muchmore dynamic than dedicated content providers. Third, the quality of service (QoS) of end-users is difficult tobe guaranteed due to heterogeneous user device and network conditions.In this project, we will work with our industrial partner and develop a cloud-assisted solution for crowdsourcedlive streaming system by taking full advantage of cloud computing and video coding technologies. First, wewill design a generic cost-effective framework for crowdsourced live streaming to support large scalegeo-distributed video crowdsourcers. Second, the heterogeneity of both broadcasters and the viewers are takeninto consideration to design transcoding and bit-rate adaption strategies for the system. Third, rate control andplay buffer optimization will be employed to further improve the QoS of end-users. Last, we will explore ahybrid cloud and peer-to-peer system as a possible approach for improving performance of crowdsourced livestreaming.