User-centric Video in the Future Internet: QoE in Participatory Video Generation and Distribution

Today, more than 73 percent of all transmitted data on the Internet is video traffic, making it the central network application which is used by billions of users globally; with new service offerings, improving content quality and, an increasing number of customers, also, new challenges arise in this domain. For example, streamed video is viewed and shared more than ever on mobile devices, bandwidth requirements rise to support standards with superior qualities like 4K and HDR, and worldwide service offerings come with diverse network environments to handle. Driven by these challenges, this dissertation presents research with the central goal to measurably improve users’ Quality of Experience in current and future video applications on the Internet. We present findings in integral parts of video streaming applications, comprising adaptive live mobile broadcasting and video on demand use cases within three integrative research areas. In our first contribution, we initially present results of a measurement study on live mobile video broadcasting services that show the video upload quality to be particularly impaired when mobile connections are used. For the automatic composition of live video, the quality of such mobile broadcasts are a prerequisite for achieving a high user satisfaction by switching between the best available content from multiple sources. However, the current approach to upload all available live user-generated video streams for mobile video composition leads to a high overhead on mobile devices. Our work presents a new method based on device context measurements that allows to drastically improve efficiency in such automatic video composition systems by identifying the relevant quality indicators on the device based on derived sensor and network measurements. We achieve an improved Quality of Experience with our proposed context-based stream selection method as verified in a field test and a crowd-sourced user study. Next, in the context of the distribution of video on demand content using Dynamic Adaptive Streaming over HTTP (DASH), we show that strong potential lies in investigating the cross-layer configuration space of video streaming systems, given the wide range of interdependent system aspects, environments, and service requirements as opposed to state-of-the-art research that focuses on single system aspects such as adaptation algorithms. By generating a broad set of experiments, i. e., covering a wide spectrum of cross-layer DASH video streaming system configuration parameters, we identify such performance aspects related to, e. g., the TCP congestion control, adaptation algorithms, and DASH players within heterogeneous network environments. We show that a subset of concrete configurations can improve DASH user experience in video on demand applications, and further motivate transitions of such DASH mechanisms based on learned sweet spot configurations. Last, we envision that in the long term, more fundamental changes to the underlying network infrastructure of the Internet need to be considered for addressing the demands of developing video streaming systems by investigation of adaptive video distribution in Named Data Networks (NDNs). First, we show that the naïve application of established concepts in DASH adaptation algorithms, that use buffer or segment throughput measurements as input, lead to unfavorable results given substantial differences in the network behavior of NDN. Our proposed concept for adaptation algorithms in NDNs is based on an improved network throughput measurement method and is shown to reduce stalling and increase streaming bitrates as compared to approaches used in current DASH adaptation algorithms. Overall, this dissertation provides the following contributions: i) first, a detailed emulation-based analysis and comparison of today’s DASH system implementations and algorithms, ii) novel concepts to enable efficient live mobile video composition, iii) and last, significant improvements in the performance for adaptive video streaming systems with the emerging NDN paradigm