User-Perceived Quality-Aware Adaptive Delivery of MPEG-4 Content Nicola Cranley Dept. of Computer Science University College Dublin Belfield, Dublin, Ireland Nicola.Cranley@ucd.ie Liam Murphy Dept. of Computer Science University College Dublin Belfield, Dublin, Ireland Liam.Murphy@ucd.ie Philip Perry School of Electronic Engineering, Dublin City University Glasnevin, Dublin, Ireland PerryP@eeng.dcu.ie ABSTRACT Many adaptive delivery mechanisms have been devised for streaming multimedia over best-effort IP networks. Most of these adaptive schemes do not consider the user’s perception of quality when making adaptations. We propose that an optimum adaptation trajectory exists which indicates how encoding quality should be adapted (upgraded/downgraded) with respect to user perceived quality in response to network conditions. This optimum adaptation trajectory can be used with any transmission adaptation policy. We describe a system architecture that uses knowledge of user perceived quality to make adaptation decisions and give an example of how this knowledge can be used to complement the sender-based adaptation algorithm, LDA. Categories and Subject Descriptors J.7 [Computer Applications]: Computers in Other Systems, Real time General Terms Algorithm, Human Factors, Performance Keywords Quality, Perception, Multimedia, Adaptation algorithm 1. INTRODUCTION The concept of adaptive video streaming is based on the widely accepted maxim that users prefer a reduced bit rate to packet losses [1]. Most of the adaptive delivery mechanisms that have been devised for streaming multimedia content do not consider user-perceived quality when making adaptations. The Real-time Transport Control Protocol (RTCP) was developed to provide a feedback mechanism to enable such adaptation over IP networks [2]. However, the network-level QoS parameters provided by RTCP (packet loss, round trip delay, and jitter) are not directly related to video content nor do they provide an accurate indication of the user- perceived quality of the played out stream. Although RTCP feedback can be used as a coarse indication of user-perceived quality, the use of application-specific feedback can enhance the accuracy of the adaptation. In general, adaptation policies (including sender- based, receiver-based and encoder-based schemes) address the problem of how to adapt in terms of adjusting the transmission rate or the window size. For example, the sender-based scheme Loss-Delay Adaptation Algorithm (LDA) [3] and its variants [4, 5] indicate how to adjust the transmission rate of the sender in response to various network conditions of loss and delay. But for a given bit rate, there are several ways to encode the content. Similarly, the receiver-based scheme RLM (receiver- driven layered multicast [6]) devises a policy of join/leave experiments for additional layers of video to improve the perceived quality and make the most of the available bandwidth. Again, the base and enhancement layers can be composed in a large number of different ways. In this paper, an adaptive streaming system is proposed that uses knowledge about user-perceived quality as the basis for adapting in response to changes in network conditions. Our work is based on the hypothesis that within the set of different ways to achieve a target bit rate, there exists a combination that maximizes the user-perceived quality. It follows that an adaptive server should modify its streaming process to attempt to find a set of encoding parameters that is close to this maximum. This paper is structured as follows. Section 2 proposes that an Optimum Adaptation Trajectory (OAT) exists for each class of video content and suggests how these trajectories could be determined. Section 3 describes an adaptive streaming system that uses RTCP feedback to infer the user-perceived quality and uses this knowledge to make adaptation decisions. Section 4 demonstrates how knowledge of the OAT can be integrated into any sender-based adaptation algorithm and, as an example, we show how this system would operate in practice with LDA. Some conclusions are presented in Section 5. Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. NOSSDAV’03, June 1-3, 2003, Monterey, California, USA. Copyright 2003 ACM 1-58113-694-3/03/0006…$5.00.