Towards Highly Interactive Networked Multimedia Services with Crowd Intelligence

随着社交、移动的和云计算的进步，众包已经成为当今数字媒体制作和共享的规范，其利用在线社区中的用户的集体努力。毫无疑问，未来的网络服务将侧重于用户体验和来自人群的丰富媒体参与。然而，社交媒体共享的无限覆盖带来了前所未有的规模挑战。高度多样化的内容类型、来源和分发渠道进一步增加了个体参与者之间的复杂交互，特别是对于高级沉浸式应用，例如交互式直播/游戏直播、多方媒体制作、在线游戏和增强现实（AR）。他们对质量和响应能力的要求要高得多（例如，延迟大约为100 ms，或者对于AR甚至为10 ms）。** 这些丰富的互动，另一方面，提供丰富和低成本的数据，以前是不可用的。利用先进的数据分析和学习工具，可以将人群中隐藏的智能揭示出来，以实现更好的系统设计。例如，从实时聊天消息中推断游戏进度，并通过注视模式跟踪视口，以进行资源预留和显示优化。然而，未经校准的数据在野外可能非常嘈杂，更不用说互联网的内在异质性和动态性了。我们的研究以以下关键问题为指导：如何在人群中有效地生成和交付大量数字媒体内容，并进行密集的在线互动和智能化？我们的长期目标是系统地研究先进的互动和众包多媒体服务的挑战，并开发集成的解决方案，将容纳和利用的互动。在解决方案框架内将解决一系列问题：（1）众包交互数据的在线分析;（2）高级计算和通信架构，特别是延迟敏感应用的云边缘协作;（3）超低延迟和低能耗数据传输。除了简单地将各个结果放在一起，在系统集成期间将重新审视各个模块的设计。这种迭代过程最终将导致优化的连贯设计，使未来的多媒体服务敏捷，高效和智能，从长远来看，有利于新兴的应用程序，如辅助/自动驾驶。我的团队在最先进的计算和通信系统方面的广泛专业知识使我们为探索这些新的研究前沿做好了准备。我们将培养高素质的人才，并将培养与世界各地的学术研究人员以及与当地和国际工业伙伴的广泛合作，以寻求技术转让，从而提高加拿大在这一优先领域的全球形象。

With the advances in social, mobile and cloud computing, crowdsourcing has become the norm in today's digital media production and sharing, which utilizes the collective effort of users in the online community. With no doubt, networked services in the future will focus on the user experience and participation with rich media from the crowd. The boundless coverage of socialized media sharing however presents unprecedented scale challenges. Highly diversified content types, origins, and distribution channels further impose complex interactions among the individual participants, particularly for advanced immersive applications such as interactive livecast/gamecast, multi-party media production, online gaming and augmented reality (AR). Their demands on quality and responsiveness are much higher (e.g., latency in the order of 100 ms, or even 10 ms for AR) than conventional IPTV or VoIP. ******These rich interactions, on the other hand, provide rich and low cost data that were previously unavailable. Leveraging advanced data analytics and learning tools, the hidden intelligence from the crowd can be unveiled toward better system design. For instance, inferring game progress from live chat messages and tracking viewport through gaze patterns for resource reservation and display optimization. Yet uncalibrated data in the wild can be highly noisy, not to mention the intrinsic heterogeneity and dynamics of the Internet.******Our research is guided by the following critical question: How can massive digital media content be effectively generated and delivered among the crowd with intensive online interactions and with the intelligence therein? Our long-term objective is to systematically examine the challenges in advanced interactive and crowdsourced multimedia services and to develop integrated solutions that will accommodate and utilize the interactions. A series of issues are to be addressed within the solution framework: (1) Online analysis of crowdsourced interaction data; (2) Advanced computing and communication architecture, in particular, cloud-edge collaboration for delay-sensitive applications; (3) Ultra-low-latency and -energy data transport. Beyond simply putting the individual results together, the design of individual modules will be revisited during system integration. This iterative process will eventually lead to an optimized coherent design, making the future multimedia services agile, efficient, and intelligent, and, in the long run, benefiting emerging applications without human-in-the-loop, such as assisted/autonomous driving.******My team's broad expertise with state-of-the-art computing and communication systems has prepared us for exploring these new research frontiers. We will train highly qualified personnel and will nurture extensive collaboration with academic researchers worldwide, as well as with local and international industrial partners to pursue technology transfers, so as to raise Canada's global profile in this priority area.**