Vicky Hardman

A survey of packet loss recovery techniques for streaming audio

We survey a number of packet loss recovery techniques for streaming audio applications operating using IP multicast. We begin with a discussion of the loss and delay characteristics of an IP multicast channel, and from this show the need for packet loss recovery. Recovery techniques may be divided into two classes: sender- and receiver-based. We compare and contrast several sender-based recovery schemes: forward error correction (both media-specific and media-independent), interleaving, and retransmission. In addition, a number of error concealment schemes are discussed. We conclude with a series of recommendations for repair schemes to be used based on application requirements and network conditions.

Successful multiparty audio communication over the Internet

The Internet was once perceived as a computer network used by researchers to transfer files and send text messages. Today, more users are becoming aware of its potential as a general communication network. Multicast conferencing over the Internet has the potential to offer low-cost real-time multimedia solutions to a wide range of user groups, provided that sufficient audio quality can be sustained. C ommercial interest in Internet audio has focused primarily on point-to-point applications such as Internet telephony, which provides roughly the same functionality as Public Switched Telephone Networks (PSTNs) over a computer network. The second focus of Internet audio developers has been download-ing audio files—typically from a WorldWide Web server—for playout on a remote users workstation [9]. Multicast conferencing [1], on the other hand, allows real-time multiway audio and video communication over the Internet and is now moving from the pilot stage [7] to a usable service in countries like the U.K. and the U.S. Multicast audio allows groups of users to participate in real-time, simultaneous audio conferences, supporting communication that goes beyond the possibilities of telephony or broadcast technology. Since the multicast backbone (Mbone—an overlay over the Internet [7]) can also support video and shared workspace, collaboration environments can be tailored to support the requirements of many distributed user groups. Another important benefit, particularly for applications such as distance education, is that multicast conferencing costs a fraction of the cost of other solutions. While video and shared data are essential to many distributed tasks, audio of sufficient quality is a necessary condition for almost any successful real-time interaction. Therefore, ensuring sufficient audio quality is a major stepping stone for realizing the potential of multicast conferencing.

Lip synchronisation for use over the Internet: analysis and implementation

This paper presents the first implementation of multicast interstream synchronisation over the Mbone/Internet. Variable bit-rate video, packet audio with silence supression and the unpredictable Mbone traffic characteristics provide a real test for the design. The paper also describes an efficient novel video tool architecture to provide intra-stream synchronisation, and implementation of interstream synchronisation using a local conference bus. Subjective performance results indicate that the efficient implementation is good enough to provide lip synchronisation for multimedia conferencing applications.

Skew detection and compensation for Internet audio applications

Long lived audio streams, such as music broadcasts, and small differences in clock rates lead to buffer underflow or overflow events in receiving applications that manifest themselves as audible interruptions. We present a low complexity algorithm for detecting clock skew in network audio applications that function with local clocks and in the absence of a synchronization mechanism. A companion algorithm to perform skew compensation is also presented. The compensation algorithm utilises the temporal redundancy inherent in audio streams to make inaudible playout adjustments. Both algorithms have been implemented in a simulator and in a network audio application. They perform effectively over the range of observed clock rate differences and beyond.

The multicast multimedia conference recorder

The ability to archive multimedia data is required in most applications of multimedia conferencing. This paper describes the generic client-server architecture of a multicast multimedia server system and the design and implementation of the multicast multimedia conference recorder (MMCR). The MMCR is a remotely controlled system capable of recording and playing back multimedia data over the MBone. It has a client-server architecture and the logical independence between components simplifies development and component replication. The MMCR uses indexes to facilitate fast data access, allowing efficient fast-forward, rewind and random access operations. Inter-stream synchronisation between recorded audio-video streams is improved by eliminating the network delay variance.

Coding of animated 3-D wireframe models for internet streaming applications

In this paper we present a coding technique for 3-D animated wireframe models, suitable for Internet streaming. First we present the MPEG-4 like coding scheme with simple RTP packetisation, and provide details of the achieved compression efficiency. Then, we describe a distortion metric for such a signal and we conduct experiments to study the effect of packet loss, following a bursty packet loss model that approximates a simulated IP network. Our experiments examine the performance of the coding scheme for a simple streaming scenario with different sequence configurations. The results show that short-term and short average length burst losses of up to 30% have a logarithmic effect on the decrease of the animation smoothness in the case of simple differential coding. This logarithmic decrease is corrected to linear by inserting I-frames to the sequence at the expense of reduced compression. The result is smoother animation.

Content-Aware Quality Adaptation for IP Sessions with Multiple Streams

While a considerable amount of research has been conducted to address QoS issues for best-effort Internet multimedia applications by utilising network-centric metrics (loss, delay, RTT, available bandwidth), less attention has been paid to the quality that is perceived by the users of the networked applications. Perceived quality of encoded multimedia is highly dependent on the time-varying characteristics of the content. We describe an approach for content-aware quality adaptation of multimedia sessions consisting of an ensemble of concurrent flows relevant to the presentation scenario. Using a quality metric that is based on the properties of the human visual perception process, we devise mechanisms that improve the overall session quality by efficiently apportioning the session bandwidth to the participating flows at appropriate adaptation times. We discuss the approach, propose suitable adaptation time scales and present results from trace-driven simulations that show the potential of content-aware quality adaptation.

Utility-Based Inter-stream Adaptation of Layered Streams in a Multiple-Flow IP Session

Future multimedia applications will evolve to content-rich, interactive presentations consisting of an ensemble of concurrent, related to the presentation scenario, flows. Recent research highlights the importance of co-ordinating adaptation decisions among participating flows in order to share common congestion control state. We exploit models that quantify the effects of the dynamics of hierarchically encoded multimedia content on perceived quality and present a mechanism to apportion the sessions aggregate bandwidth among its streams that improves the total session quality. Dynamic bandwidth utility curves are introduced to express the variability of multimedia content and represent the level of quality (or satisfaction) an application/user receives under given bandwidth allocations. The relative importance of the participating flows, determined either by the user or the application scenario, is also considered. We discuss our approach and analyse simulation results obtained based on trace-driven simulation.

Improving the Quality of Recorded MBone Sessions Using a Distributed Model

Multicast conference recording currently uses a single, arbitrarily placed recorder. The recordings have less than perfect quality because they are often not loss-free, and the recording is as perceived by the recorder (delays from senders to the recorder have been imposed on the streams). An ultimate recording consists of the loss-free data set, extra information about loss patterns received by individual participants, and information about the delays between participants. The delay information is required so that we can re-create the conference from a number of perception points. In this paper, we initially consider off-line recording of material, which produces loss-free recording, can provide timing information, and is simple, but does suffer from deficiencies. Our second (and preferred) solution uses a distributed arrangement of a number of co-operating recording caches, and sender interaction far local repair.