2.6 Chapter Summary
3.2.3 User Experience-aware Solutions
In [156], the authors present a user-perceived quality aware adaptive scheme for MPEG- 4 delivery over best-effort IP networks. The proposed adaptive scheme is using an optimum adaptation trajectory (OTA) where 120 subjective tests were set up to ask users to select their preferred clips. A fitting curve is then built to model the relationship between resolutions and frame rate. The OTA curve is used to adjust the quality level of the streamed videos in the adaptive server, which maximizes the perceived video quality by the end-users. However, the adaptive scheme has not been implemented in a real streaming server and validated in second-round subjective tests after video streams were received at the client side.
A QoE-aware DASH system (QDASH) is proposed in [157] which consists of QDASH- abw and QDASH-qoe functionalities. QDASH-abw is an available channel bandwidth measurement system detecting the highest quality level of video that the current net- work conditions can support. QDASH-abw makes use of a probing methodology to send and collect the probing packets via a measurement proxy. It is an accurate method to determine the sending rate and available bandwidth via the probing packets samples during Round-Trip-Time (RTT). On the other hand, QDASH-qoe adapts the streaming rate depending on the buffering information feedback via the DASH standard and the available bandwidth estimated by QDASH-abw.
Essaili et al. in [158] propose a Quality of Experience (QoE) driven algorithm for multi-user over Dynamic Adaptive Streaming over HTTP (DASH) in LTE networks. In order to enhance QoE awareness to the end-user, the proposed scheme makes user of an QoE estimation model which calculate the Mean Opinion Score based on the video stream data rate variation. Jointly considering the estimated MOS of each video and channel conditions and for each end-user the optimal streaming rate is adjusted. This QoE driven scheme also implements a QoE-based proxy server which is redirecting the video requests from the end-users and proactively adapts the streaming rate while the channel condition is bad. The proposed solution was evaluated and was implemented for two standard adaptive HTTP clients: Microsoft Smooth Streaming and DASH-enabled VLC.
NOVA (Network Optimization for Video Adaptation), a simple asymptotically op- timal online algorithm for QoE-driven DASH-based video delivery, is proposed in [159].
NOVA is a joint optimization system between network resource allocation scheme and video quality adaptation scheme. In order to maximize the perceptual video quality of the end-user with a trade-off among the mean quality, temporal variability in quality and fairness. Distributing the resource allocation to network controllers and adaptation tasks to end-users, NOVA is asynchronous and more flexible, which is suited for current network and normal DASH system.
Many QoE-aware adaptive schemes focus on how to adapt a proper stream to the end-users switching between the stream bitrates. However, the frequent switching of the bitrate from high to low or low to high would cause higher streaming overhead and worsen the user experience. Moreover in [160], the authors present a QoE friendly algorithm to solve this. It makes use of a fixed-interval buffer to maintain the unchanged bitrate and to keep a smooth switching-up or switching-down. Additionally, once the buffer is going to overflow, then a quick boot algorithm is exploited to get the proper bitrate to fitting a current bandwidth.
A QoE-driven adaptive streaming algorithm formulated from real subjective exper- iments results is presented in [161]. In order to maximize the QoE to the end-users, the proposed algorithm investigates the problem of how to cache and manage a set of media files with optimal streaming rates. Then, the authors formulate this problem as a convex optimization problem which is going to be solved with Lagrange multi- plier method. Through three alternative search algorithms (i.e. exhaustive search, dichotomous-based search and variable step-size search), the authors find the optimal number of cached media files for high expected QoE and low complexity. However, this algorithm of adaptive streaming cache management is only evaluated in a single content scenario. Whereas multiple distinctive contents stored on a cloud-based server would be considered as part of the future work.
A cross-layer Video-QoE adaptive scheduling scheme jointly working with DASH is presented in [162]. The proposed scheme based on the Gradient Algorithm makes use of the periodic buffer level feedback (i.e. standardized in DASH standard) of user-clients to prioritize the streamed user at each time instant. Also the scheme considers the CQI feedback on the scheduling process. The performance evaluation shows that the scheme is able to provide the continuous adaptive streaming rate for user-clients with a good perceptual video quality (i.e. computed by the difference between the mean and
standard deviation of PSNR) compared to the other three scheduler: conventional PF scheduler, PF with barrier for frames (PFBF) scheduler [163] and Gradient with Min rate (GMR) scheduler [164]. Additionally, the scheme could be flexibly exploited in non-video streaming service by choosing custom optimization criterion (i.e. changing buffer level feedback to delay or loss tolerance).