Joint source-channel coding as an element of a QoS framework for ‘4G’ wireless multimedia

https://doi.org/10.1016/j.comcom.2003.10.016Get rights and content

Abstract

The development of ‘Beyond 3G’—B3G—or ‘4G’ networks and applications, with all-IP based ubiquitous and seamless service provisioning across heterogeneous infrastructures, presents a number of challenges beyond existing capabilities conceived so far by the IETF for Mobile IP and by the ITU for Third Generation networks. Management, signaling and transport functions must evolve from different link layer technologies, uniformly and end-to-end, to the packet switched IP-network layer. Global mobility throughout all network types requires capability for seamless handover between different administrative and technology domains, extending at the same time the present edge mobility concept to ad-hoc networks and to mobile networks. Having at least identified, if not solved, these important problems, the technical key challenge for this B3G or ‘4G’ world remains to provide IP-based audio and video communications at the same or at least comparable ‘carrier class’ and bandwidth efficiency as the corresponding circuit-switched subsystems of 3G and 2G. In the IETF, some elements to achieve this goal have been identified and are under discussion: situated between link and IP layers, the Robust Header Compression (ROHC) framework addresses the problems of spectrum and bandwidth scarcity and of bit errors characterizing wireless links. The fact that, for some applications, erroneous packet payloads can be valuable and better to cope with than lost ones, has inspired the introduction of the UDP-lite transport protocol. Assuming ROHC, UDP-lite and an appropriate inter- or cross-layer signaling mechanism in action, this paper is going to elaborate and incorporate the joint source and channel coding—JSCC—paradigm into a complete QoS framework for 4G mobile wireless multimedia communications.

Introduction

All-IP 4G networks are only a vision so far. Without any official definition, most related work attributes the following elements to a 4G world: significantly higher data rates and cell capacities compared to 2.5 and 3G networks, worldwide roaming capabilities with vertical and horizontal handover between technologies and administrative domains, enhanced mobility forms such as adhoc and mobile networks and—possibly most important—an all-IPv6 data and control framework. While the development of more efficient new modulation and media access schemes is a broad research field of its own, and basically orthogonal to the all-IP paradigm, the all-IP and Mobile IP architecture is likely to impact and transform many of the functions both of the wired Internet and of the classical 3G world. Moving many management, control and transport functions from different link layer technologies of 3G networks and e.g. 802.11 LANs, in an efficient way uniformly and end-to-end, to the packet switched IP network layer is a major challenge.

Foremost, to provide all-IP-based multimedia, especially classical voice communication, at the same or at least comparable ‘carrier-class’ and bandwidth efficiency as the corresponding circuitswitched subsystems of 2G and 3G remains an issue to be solved. QoS provisioning becomes all the most challenging in a global mobility context with highly varying channel characteristics (bandwidth, throughput, error rates, fading and erasure characteristics…). To best meet heterogeneous sources requirements, different service classes built upon link layers making use of different transmission modes (i.e. with different channel codes, modulation and access methods), are very likely to be defined, as already present in 3G systems [1] (e.g. the UMTS layered QoS classes, conversation class, streaming class, interactive class, background class). These end-to-end services are very likely to be provided by a layered architecture, with QoS classes parameterized by sets of QoS attributes with different applicability and values in the different layers, leading in turn to end-to-end channels with different bandwidth, delay, error and loss characteristics.

One critical component of the end-to-end service is the so-called radio bearer service with key parameters such as delivery of erroneous SDUs, residual error rates, transfer delay. In a mobile environment, channel error rates up to 10% are quite common. With interleaving and error correcting codes, higher channel performances can obviously be achieved, however at the expense of delays and/or throughput reduction not compatible with some multimedia application requirements. The error rate seen by higher layers can also be reduced through local retransmissions at the link layer (use of ARQ mechanisms, Automatic Repeat Request), however again at the expense of extra delay and of a reduction of the throughput available to higher layers. Concurrently, the QoS requirements of multimedia applications are in sharp contrast with requirements of traditional computer communications. They can tolerate errors and erasures but they are highly delay sensitive.

While the Internet protocols have been designed implicitly assuming that bandwidth, delay, error rates do not vary much in the link layer, 3G systems have been aware of the varying channel characteristics of the radio bearer caused by changing radio conditions in mobile environments. Not adhering however to the Internet end-to-end and layer-separation principle, in 3GPP from the very beginning, the bearer layer always contained several error control mechanisms which could be parameterized eventually according to the needs of the application [1]. In the Internet, the strict application of the end-to-end and layer separation principle is now also discussed and questioned in the light of characteristics of the radio bearer. In Ref. [2], it is explicitly stated that ‘if a subnet contains error control mechanisms (retransmission and/or Forward Error Correction—FEC), it should be possible for the IP layer to influence the inherent tradeoffs between uncorrected errors, packet losses and delay. These capabilities at the subnet/IP layer service boundary correspond to the selection of more or less error control and/or to selection of particular error control mechanisms within the subnetwork.’ Such ideas of inter-layer communication which would allow to best select and adapt subnet technologies to varying conditions are progressing in the community [3].

Accordingly, both in the 3G and ‘all-IP-world’, it seems now to be a common understanding that the provisioning of QoS for multimedia applications such as video or audio does require a loosening and a re-thinking of the end-to-end and layer separation principle [4]. The Robust Header Compression (ROHC) framework [5], and the new UDP-lite protocol—possibly delivering erroneous packets [6]—are strong steps in that direction. The 3G system specifications as well as the UDP-lite protocol and the RHOC [5], [7] initiatives acknowledge that to achieve high utilization of scarce wireless resources, transmission modes offering guarantees in delay, possibly in bit rate, but no fixed guarantee in error rates, are required.

In this context, this paper describes the joint source-channel coding paradigm, evolving at the application layer, as an element of a QoS framework for 4G wireless Multimedia. In this framework and supported by appropriate signaling, UDP-lite and ROHC, the application layer JSCC element co-operates vertically especially with the radio link layer.

So far, the design of source coding systems has been mainly driven by objectives of compression or rate efficiency, without taking into consideration possible degradations induced by the transport of the compressed streams over networks. In contrast, the concept of joint source-channel coding is guided by an optimum trade-off between compression efficiency and error and/or erasure resilience depending on the link characteristics. In presence of errors and/or erasures, rather than trying to best approach the entropy of the source by removing all redundancy and correlation, a controlled amount of redundancy has to be maintained in the compressed source representation. Symbol or bit level forms of redundancy can be introduced at the different canonical stages of a compression scheme (transform, quantization or entropy coding) in order to fight against the different types of channel impairments. Adaptive multimedia application control making use of congestion control mechanisms, coupled with efficient fine grain scalable source representation, for dynamic rate adaptation to—possibly rapidly—varying channel bandwidth is also a key component of this framework.

The paper is organized as follows. Section 1 uses the 3GPP traffic classes and QoS attributes as an example of context in which the JSCC paradigm promoted here can already apply. It also describes the present activities of the IETF with respect to QoS in a mobile and wireless environment to give an overview of the global evolving context. After reviewing the main limitations of the traditional model of separation between source coding and delivery in Section 3, we proceed with the description of a set of ongoing JSCC research directions. Section 4 describes linear transforms, which can be regarded as channel codes on the real field, allowing to represent the source signal (audio, image, video) in a way that would allow to flexibly and efficiently trade compression efficiency against erasure recovery capability. In addition, channel diversity can be exploited by designing appropriate transmission and/or packetization schemes. Section 5 outlines several error-resilient coding and decoding solutions. Finally, Section 6 describes the integration of the JSCC framework into an overall all-IP 4G QoS architecture providing for a flexible, scalable and at the same time error and erasure resilient representation of video signals.

Section snippets

QoS for multimedia in a 4G world: related work

In this section we are going to review, bottom up, some typical and relevant elements in the present ITU and IETF context which will eventually contribute to the core of a 4G QoS framework in support of applications such as video and audio.

Traditional transform coding: issues WRT wireless channels

In this section, we first briefly review the design principles of the main components of traditional media (audio, image, video) compression systems. We then raise a few limitations concerning the delivery of corresponding compressed streams over wireless networks.

Erasure resilient source coding

Let us first consider the problem of erasures. In this type of degradation, the receiver either receives the message (packet) correctly or does not receive it at all. In addition, it knows the locations of the erased samples (the lost packets). Compression standards, such as MPEGx and H26x, support mechanisms such as prediction modes restriction, coding mode selection taking into account the signal distortion induced by the erasures [13], [14]. The adaptation of coding modes to the loss

Error resilient source coding

Entropy coding, producing VLC, is a core component of any data compression scheme. The mostly used variable length codes are Huffman and arithmetic codes. Huffman codes are constructed according to the stationary probability distribution Ps of the source of symbols. If Ps is a dyadic law, then the Huffman coder achieves Shannon's compression bound: the average length of a codeword is equal to the entropy of Ps. As a consequence, the bitstream at the output of the coder is composed of close to

Steps towards a complete joint source-channel coding system for 4G wireless multimedia: example of video

In order position the JSCC paradigm as an element of a complete QoS framework for 4G wireless multimedia, in this section we are going to add missing elements and to integrate the overall system in a coherent way. The exemplified and partly new JSCC components provide both for flexibility with respect to variations in radio link quality and, most important, also for backward compatibility with ‘non-redundant’ decoders. We first present the HAN signaling framework in the context of a

Conclusion

The vision of all-IP 4G networks, before becoming reality, still requires significant research efforts in various orthogonal and complementary directions. In this paper, we have tried to give an overview of a path towards QoS for multimedia in a mobile and wireless all-IP 4G world. We have outlined a complete ‘content and channel adaptive’ QoS system expected to be beneficial in this context. Such a system can be eventually complemented by classical QoS architectures derived from present IETF

References (68)

  • V.K. Goyal et al.

    Quantized frame expansions with erasures

    Appl. Comput. Harmonic Anal.

    (2001)
  • Third generation partnership project, Technical specification group services and system aspects; QoS concept and...
  • P. Karn, Advice for internet subnetwork designers, draft-ietf-pilc-link-design-12.txt, IETF working group on...
  • L.-A. Larzon, U. Bodin, O. Schelen, Hints and notifications, in: IEEE International Wireless Communications and...
  • J. Kempf, R. Austein, A quality-of-service resource allocation client for casp., Internet Draft...
  • C. Bormann, C. Burmeister, M. Degermark, H. Fukushima, H. Hannu, L.-E. Jonsson, R. Hakenberg abd, T. Koren, K. Le, Z....
  • L.-A. Larzon, M. Degermark, S. Pink, L-E. Jonsson, G. Fairhurst, The UDP-lite protocol,...
  • L-E. Jonsson, G. Pelletier, Robust header compression (ROHC): a link-layer assisted profile for IP/UDP/RTP, IETF, RFC...
  • Internet Architecture Board, IAB wireless internetworking workshop, http://www.iprg.nokia.com/hinden/IAB-workshop/,...
  • H. Chaskar, Requirements of a QoS solution for mobile IP, draft-ietf-mobileip-qosrequirements-03.txt, IETF Mobile IP...
  • G. Pelletier, Profiles for UDP-lite protocol, Internet draft draft-ietf-rohc-udp-lite-00.txt, IETF Network Working...
  • L.A. Larzon, Three problems with internetworking in cellular networks, PhD thesis, Lulea University, Sweden, September...
  • H. Schulzrinne, Profiles for UDP-lite protocol, Internet draft draft-ietf-rohc-udp-lite-00.txt, IETF NetworkWorking...
  • T. Wiegand et al.

    Rate-distortion optimized mode selection for very low bit rate video coding and the emerging H.263 standard

    IEEE Trans. Circuits Syst. Video Technol.

    (1996)
  • R.O. Hinds et al.

    Joint block-based video source/channel coding for packet-switched networks

    (1997)
  • H. Bolcskei, F. Hlawatsch, H. Feichtinger, Frame-theoretic analysis and design of oversampled filter banks, in:...
  • R.E. Blahut

    Transform techniques for error control codes

    IBM J. Res. Dev.

    (1979)
  • T.G. Marshall

    Coding of real-number sequences for error correction: a digital signal processing problem

    IEEE J. Selected Areas Commun.

    (1984)
  • F. Marvasti et al.

    Efficient algorithms for burst error recovery using fft and other transform kernels

    IEEE Trans. Signal Process.

    (1999)
  • G. Rath et al.

    Performance analysis and recursive syndrome decoding with dft codes for bursty erasure recovery

    IEEE Trans. Signal Process.

    (2003)
  • G. Rath et al.

    Error and erasure correction using overcomplete frame expansions

    (2003)
  • J. Kovacevic et al.

    Filter bank frame expansions with erasures

    IEEE Trans. Information Theory

    (2002)
  • P.L. Dragotti et al.

    Quantized oversampled filter banks with erasures

    (2001)
  • S. Mehrotra, P.A. Chou, On optimal frame expansions for multiple description quantization, in: Proceedings of the IEEE...
  • A. Gabay, P. Duhamel, O. Rioul, Spectral interpolation coder for impulse noise cancellation over a binary symmetric...
  • R. Motwani et al.

    Tree-structured oversampled filter banks as joint source-channel codes: Application to image transmission over erasure channels

    IEEE Trans. Signal Process.

    (2003)
  • V.K. Goyal

    Multiple description coding: compression meets the network

    IEEE Signal Process. Mag.

    (2001)
  • L. Ozarow

    On a source coding problem with two channels and three receivers

    Bell Syst. Technol. J.

    (1980)
  • A.A. El Gamal et al.

    Achievable rates for multiple descriptions

    IEEE Trans. Information Theory

    (1982)
  • V.A. Vaishampayan

    Design of multiple description scalar quantizers

    IEEE Trans. Information Theory

    (1993)
  • V.A. Vaishampayan et al.

    Multiple description vector quantization with lattice codebooks: design and analysis

    IEEE Trans. Information Theory

    (2001)
  • W. Jiang, A. Ortega, Multiple description coding via polyphase transform and quantization selection, in: Proceedings of...
  • W.M. Lam et al.

    Self-synchronizing variable-length codes for image transmission

    (1992)
  • Y. Takishima et al.

    Reversible variable length codes

    IEEE Trans. Commun.

    (1995)
  • View full text