Wireless Systems for Unicast/Multicast Video Streaming Services Using
Scalable Video Coding
Daeyeon Kim
Department of Electronic Engineering The University of Electro-Communications
A thesis submitted for the degree of Doctor of Philosophy
March 2013
Studies on Resource Allocation of Wireless Systems for Unicast/Multicast
Video Streaming Services Using Scalable Video Coding
APPROVED BY SUPERVISORY COMMITTEE
Chairperson: Associate Professor Takeo Fujii
1. Member: Professor Yoshio Karasawa
2. Member: Professor Masahide Kaneko
3. Member: Professor Takeru Hashimoto
4. Member: Professor Toshiharu Kojima
Copyright ⓒ 2013 by Daeyeon Kim
All Rights Reserved
i
無線システムでのユニキャスト／マルチキャスト ビデオストリーミングのための
資源割当に関する研究
金 大淵
論文の概要
スマートホンやタブレット PC など携帯ディバイスの急激な発展により、モバイルマ ルチメディアサービスが一般化されてきた。これに伴って、無線チャネルによるビデ オデータのストリーミング伝送に関する研究が注目されている。しかし、圧縮ビデオ ス トリー ムの場 合、一 般的 に要求 ビット レート が非 常に高 い ( 例えば HD （high definition）ビデオの場合、約 1 から 20 Mbps)。ビデオストリームは数個のパケッ トで構成されており、各パケットは他のパケットからの影響を受け、今までどのパケ ットが受信されたかによって、各受信パケットを復号できるかどうかが決められるた め、わずか１ビットの誤りでも全体的なビデオの品質が大幅に劣化する可能性がある。
そのため、ビデオをストリーミング伝送する際、一般的にその品質を保証することが 難しく、マージンを考え非常に広い帯域幅や高いチャネル品質が求められる。このよ うな問題を緩和するため、柔軟にビットレートおよびビデオ品質を提供する SVC
（scalable video coding）ビデオ圧縮方式が提案および標準化されている。SVC 方 式はビデオレイヤーと呼ばれる数個のサブビットストリームに分けてビットストリー ムを提供し、チャネル状態により、復号化およびイメージ化する際、それぞれが異な るプライオリティを持つビデオレイヤーを用いることが可能となる。例えば、送信電 力 、変調 、コー ディン グ、 パリテ ィチェ ックコ ード などを 制御す ることで UEP
（unequal protection）をビデオレイヤーに適用することが可能となる。また、状態 の良いチャネルで優先度の高いビデオレイヤーを選択して伝送する方式も考えること ができる。
無線通信ではこのようなビデオストリーミングに無線リソースを割り当てて配信する ことになる。これまで無線リソースへのビデオストリーミングに関しては一般的な無 線システムでの検討はあるが、より無線リソース割り当ての難しいマルチホップ通信 環境や、ユニキャスト通信とマルチキャスト通信が混在するような環境、基地局が連 携動作するような環境での研究はなされてこなかった。そこで、本論文では無線通信 チャネルでユニキャスト／マルチキャストを考慮した無線リソースの割り当て手法に 関しての検討を行い、ビデオ伝送の品質を確保したうえでの効率的な伝送手法につい
ii
(Modulation and Coding Scheme)を UEP に適応させて決定する。受信ビデオ画質は誤 り無く受信された SVC パケットのパターンによって決められ、各受信パケットのパタ ーンはパケット誤り率（PER）によって決定される。そのため、期待受信画質を PER で表すことができる。そのうえ、 各リンクの MCS により PER と送信時間が決定され るため、最終的には期待受信画質を時間の関数で表現して凸最適化を適用することが できる。そのため複数のユーザについての期待受信ビデオ画質が最大化されるように 転送するパケットの組み合わせおよび各リンクと各パケットの MCS を決めるアルゴリ ズムを提案する。その結果、UEP を適用しない一般的な MCS 方式より期待受信画質が 改善することが確認できた。
次に、４章ではユニキャストとマルチキャストが混在する環境において、複数の周波 数チャネルでビデオ伝送を行う手法の提案を行っている。通常のユニキャストサービ スとモバイルテレビのようなビデオ伝送サービスを混在して処理可能な手法として、
LTE(Long Term Evolution)標準で定義されているような MBMS(Multimedia Broadcast Multicast Service) を 提 供 す る OFDM(Orthogonal Frequency Division Multiplexing)方式に基づくセルラーシステムを考えている。この時、通常のユニキ ャストサービスの場合、固定ビットレートを用いての伝送となり、ビデオ伝送で複数 ユーザに送信を行うような MBMS サービスの場合は、適応的にビットレートを変更し て送信する必要がある。このような二つのサービスを効率的に提供するために送信電 力を割り当てる際には、次の事項を考慮する必要がある。まず、各々のユニキャスト サービスユーザに対しては要求されるビットレートを保証する。次に、各々の MBMS を用いるマルチキャストユーザには最低限の SVC ビデオ品質を提供するための最低ビ ットレートを保証する。本章では、これらの要求を満足させたうえ、マルチキャスト ユーザに対する平均ビデオ品質を最大化するアルゴリズムの提案を行い良好な伝送性 能が確認できた。
５章では、複数の基地局からの連携送信を考慮した際に、最低品質ユーザの品質が最 大化されるように無線リソース割り当てする手法を検討し、マルチキャスト性能の向 上を図った。この研究では、4 章に続き、MBMS のサービスを用いることとし、さらに 基地局連携による性能改善を目指した MBSFN(MBMS Single Frequency Network)を用 いることとする。MBSFN は複数の基地局で構成されており、基地局同士は同一の MBMS パケットを送信することで、マルチキャストでの伝送品質を改善させる仕組みを持っ ている。マルチキャスト伝送の場合、同一のサービスを要求するユーザグループへは、
安定的に必要なサービスを提供するため、送信パラメータを最低チャネル品質のユー ザに合わせて伝送する必要がある。そこで、本研究では、基地局間電力バランスを調 節し、ユーザグループ間での最低チャネル品質を最大化するアルゴリズムを提案し、
その有効性を確認している。
iii
３章から５章の検討では、それぞれ、シミュレーションと解析による評価を行ってい る。これらの結果から、提案アルゴリズムを用いることにより、平均ビデオ品質を考 慮した無線リソース割り当てが、スループットやビット誤り率を目標とした無線リソ ース割り当てと比べて、サービス品質を大幅に向上させることを確認した。最後に６ 章で本論文をまとめる。
v
Studies on Resource Allocation of Wireless Systems for Unicast/Multicast
Video Streaming Services Using Scalable Video Coding
Daeyeon Kim
ABSTRACT
Owing to high-performance wireless systems and mobile devices, video transmission over wireless channel for applications such as interactive video games, mobile TV or video telephony have become more feasible, and hence, supporting such applications has become a necessary part of commercial broadband services. In order to efficiently provide such services and accommodate a huge number of users, modern wireless systems are equipped with transmission schemes capable of scalable bandwidth allocation so that the limited bandwidth resources are more efficiently utilized according to heterogeneous channel conditions and services.
For cooperating with the wireless systems and exploiting their potentials for video services, scalable video coding (SVC) was developed, where a single video content coded by SVC can be truncated into a number of video layers with different priority, and a part of the layers can be dropped for provision of adjustment in bit rate and video quality. Nevertheless, if the scalable resources of the wireless systems are used without sufficiently considering the properties of the SVC streams, e.g., required bit rate and video quality of each video layer, we cannot benefit from the bit rate flexibility of the SVC streams.
In this thesis, we provide several use cases of SVC for providing video services, where various wireless systems defined in fourth-generation (4G) standards, i.e., multihop, multi-channel and multi-base station (BS) systems,
ABSTRACT
vi
are assumed for transmitting the SVC streams. Specifically, we consider improving the service quality, in terms of received video quality, by adequately utilizing the bit rate flexibility of the SVC streams and wireless resources that can be manipulated owing to the transmission scheme of the systems.
In Chapter 3, we provide an error-bandwidth control problem in physical layer for transmitting the SVC streams over a path consists of a source and a destination node and multiple relay nodes by using adaptive link adaptation based on priority of the video layers. That is, using more intelligent reliability criteria for link adaptation at each link between neighboring nodes according to the relative importance of the video layers, we aim to improve expected video quality at the receiving node, which results from packet drop rate decided by the source node and packet error rates of the links.
In Chapter 4, we address another problem for video service using the SVC streams: resource allocation for a base station (BS) using orthogonal frequency division multiple access (OFDMA) scheme for unicast and multicast services, where SVC stream is considered for each multicast service. In this problem, we aim to guarantee minimum bit rates for unicast and multicast services, and to maximize the average video quality of the multicast services.
In Chapter 5, we also address an OFDMA resource allocation problem for improving the service quality of the cooperative multi-BS system. By introducing concept of power allocation ratio (PAR), we decompose the resource allocation problem into two sub-problems, PAR allocation problem among the BSs for each OFDMA subchannel (CH) and power allocation problem among the CHs. The two sub-problems can be solved independently with separate algorithms for maximizing the worst user’s channel condition and for maximizing system utility, respectively.
Based on the heterogeneous channel condition of the users and the bit rate characteristics of the SVC streams, algorithms proposed in this thesis can be used for improving the quality of the video services provided by the wireless systems in terms of not only the total bit rate but the total video quality which is directly related to the user’s satisfaction.
vii
ACKNOWLEDGEMENTS
I would like to sincerely thank my supervisor Prof. Takeo Fujii who gave me the chance to start my study in the University of Electro- Communications and also a lot of opportunities to participate on high quality technical conferences. Owing to his advice and help with patient, I could take sufficient efforts and time to motivate and complete my research.
I also would like to express my sincere gratitude to the thesis committee: Prof. Yoshio Karasawa, Prof. Masahide Kaneko, Prof.
Takeru Hashimoto, and Prof. Toshiharu Kojima for their valuable comments and suggestions which guided me to modify and prepare this dissertation.
I owe a debt of gratitude to Prof. Kyesan Lee for his guidance and giving me this great opportunity to study in Japan. I also would like to thank Prof. Doug Young Suh, supervisor of my master’s course in Kyung Hee University, for all his advises, supports and encouragements.
I also would like to thank all my friends, who always help and support me during my stay in Japan. I would like to give special thanks to my parents and my parents-in-law for their continuous supports and encouragements.
Finally, I please to dedicate this dissertation to my lovely wife Yuna and my son Yunsu for their unconditional love and support. Since they are my courage, happiness, energy, belief, and everything, they deserve thanks from my heart.
TABLE OF CONTENT
ix
TABLE OF CONTENT
Chapter 1. Introduction...……….1
Chapter 2. Overview of Resource Allocation Algorithm………...7
2.1. Scalable Video Coding.... ………...…...7
2.2. Video Transmission over Multi-hop Channels ………...………….9
2.3. Resource Allocation for OFDMA .. ...12
2.4. Resource Allocation for Multicast Service...15
2.5. Resource Allocation for MBSFN...16
2.6. Proposed Resource Allocation Algorithms... 17
2.6.1. Proposed Algorithm in Chapter 3...17
2.6.2. Proposed Algorithm in Chapter 4...18
2.6.3. Proposed Algorithm in Chapter 5...19
Chapter 3. Modulation Level Allocation for Unicast Service over Multi-hop Channels...21
3.1. Expected Video Distortion...21
3.1.1. Simple Examples of Expected Distortion...22
3.1.2. Expected Distortion of Hierarchical B Structure...25
3.1.3. Expected Video Distortion Increment...27
3.2. Time Resource Allocation Algorithm...28
3.2.1. Definition of Resource-Distortion Attribution...30
3.2.2. Continuous Modulation and PDR Setting...33
3.2.3. Parameters Refinement...37
3.2.4. Total Time Constraint...38
3.2.5. Practical Expected Distortion Information...41
3.2.6. Complexity...42
3.2.7. Discrete Modulation Setting...42
3.3. Simulation Results...43
3.3.1. Single-hop Link...43
3.3.2. Two-hop Link...50
3.3.3. Three-hop Link...52
3.4. Summary of Chapter 3...56
x
Chapter 4. Power and Bandwidth Allocation for Unicast and Multicast Service
over OFDMA Channels...57
4.1. Related Works...58
4.1.1. Resource Allocation Algorithms for OFDM-based Multicast Service...58
4.1.2. Limitations in Related Works ...59
4.2. System Configuration...62
4.3. Proposed Algorithm...63
4.3.1. Concave Approximation of Service Quality...63
4.3.2. Multicast Configuration...68
4.3.3. Resource Allocation...70
4.4. Simulation Results...79
4.4.1. Bit Rate Allocation of Proposed Algorithm...80
4.4.2. Multicast Gain of Proposed Algorithm...84
4.4.3. Performance Comparison...86
4.5. Summary of Chapter 4...89
Chapter 5. Power and Bandwidth Allocation for Multicast Service over OFDMA Channels Using Multi-BS System………..…...91
5.1. System Configuration...92
5.2. Related Work...93
5.3. Proposed Algorithm...94
5.3.1. Worst Channel Gain Maximization Algorithm...95
5.3.2. Total Utility Maximization Algorithm...96
5.4. Simulation Results...99
5.5. Summary of Chapter 5...103
Chapter 6 Conclusion...105
6.1. Contribution of Research...105
6.2. Future Research Work...107
LIST OF FIGURES
xi
LIST OF FIGURES
Figure 2.1. Coding structure of a GOP quality layers...8
Figure 2.2. Bit rate-PSNR performance...9
Figure 2.3. Slice and multiple description coding………....11
Figure 2.4. Quality of Frequency selective channel………...….12
Figure 3.1. Reference structure of three frames...23
Figure 3.2. Reference structure of two frames with two quality layers...24
Figure 3.3. Hierarchical B structure with four temporal layers...26
Figure 3.4. Reference group representation of Figure 3.3...26
Figure 3.5. vs. , according to (3.24)………..35
Figure 3.6. Algorithms for finding and for a given ………...36
Figure 3.7. Output from Algorithm II ……….….39
Figure 3.8. MSE increment………...44
Figure 3.9. Number of bits according to the number of transmitted packets…44 Figure 3.10. Number of packets transmitted in 0.2 sec……….…...…45
Figure 3.11. Transmission time and EPSNR of FBeD………...…47
Figure 3.12. MSE increment and BER levels allocated by FBeD and SLO.…48 Figure 3.13. Transmission time allocated by FBeD and SLO ……….…49
Figure 3.14. Number of transmitted packets and EPSNR………50
Figure 3.15. Transmission time for each link……..……….…51
Figure 3.16. Number of transmitted packets and EPSNR……….……52
Figure 3.17. EPSNR for various channel SNRs………...53
Figure 4.1. Position of the bits in a bit stream………….……….61
Figure 4.2. A bit stream that can be divided into two parts………..……64
Figure 4.3. Bit value according to the step function approximation…….……66
Figure 4.4. Bit value of test video “Football”……….…..67
Figure 4.5. Position of the bits in a bit stream………..…69
Figure 4.6. Power and subchannel allocation function……….74
Figure 4.7. Configuration of a BS and users in a 4 Km2square………80
Figure 4.8. Bit rates allocated to the users in Figure 4.7………..81
Figure 4.9. Bit rates of five CGs for streaming test video “Football”…….….82
xii
Figure 4.10. Average PSNR of fifteen users……….84
Figure 4.11. Average PSNR of (a) thirty and (b) fifty users……….85
Figure 4.12. Bit rates of six CGs by previous algorithm………..87
Figure 4.13. Bit rates of six CGs by modified algorithm………..87
Figure 4.14. Average PSNR of six CGs by three algorithms……..……….…88
Figure 5.1. Simulation configuration and bit rate performance………….….100
Figure 5.2. Bit rate of four streams with different utilities……….…101
LIST OF TABLES
xiii
LIST OF TABLES
Table 3.1. Possible CNEPs of the reference structure in Figure 3.1……….…23 Table 3.2. Possible CNEPs of the reference structure in Figure 3.2………….24 Table 3.3. CNEPs of the RG structure in Figure 3.4………27 Table 3.4. Calculations for to approach in Figure 3.5…………37 Table 3.5. Reduced number of CNEPs ………41 Table 4.1. Multicast patterns which can be considered for three users………60 Table 4.2. Proposed multicast patterns for four component groups………..69 Table 4.3. Proposed multicast patterns for seven CGs ………70
xiv
ACRONYMS
xv
ACRONYMS
3GPP 3rd Generation Partnership Project
AVC Advanced Video Coding
BCMCS BroadCast/MultiCast Service
BL Base Layer
BS Base Station
CGS Coarse-Grain quality Scalability CIF Common Intermediate Format CNEP Combination of NEP
CSI Channel State Information
ECH EL-transmit CH
EL Enhance Layer
FPS Frames per Second EPA Equal Power Allocation FBeD Flexible BER Decision FEC Forward Error Corrections GOP Group of Pictures
IEEE Institute of Electrical and Electronics Engineers JSVM Joint Scalable Video Model
LTE Long Term Evolution
MBMS Multimedia Broadcast Multicast Service MBS Multicast Broadcast Service
MBSFN MBMS Single Frequency Networks MCE Multi-Cell/Multicast Coordination Entity MDC Multiple Description Coding
WCG Worst Channel Gain
WCGMA WCG Maximization Algorithm MCS Modulation and Coding Schemes MI-BS Most Influential BSs
MGS Medium-Grain quality Scalability MLO Multi-Link Optimization
MQAM Multi-level Quadrature Amplitude Modulation
xvi MSE Mean Squared Error
NEP Number of Effective Packets
OFDM Orthogonal Frequency Division Multiplexing OFDMA Orthogonal Frequency Division Multiple Access PAR Power Allocation Ratios
PDR Packet Drop Rate
PER Packet Error Rates PLR Packet Loss Rate
PSNR Peak Signal to Noise Ratio
QL Quality Layers
QoS Quality of Service
RG Reference Groups
RsD Resource-Distortion SLO Single Link Optimization SNR Signal to Noise Ratio SVC Scalable Video Coding
TL Temporal Layers
TUMA Total Utility Maximization Algorithm
UMTS Universal Mobile Telecommunications System WiMAX Worldwide Interoperability for Microwave Access WLAN Wireless Local Area Network
Chapter 1. Introduction
1
Chapter 1
Introduction
Wireless communications for smart phones, laptop computers have become a part of everyday life in nowadays, which have experienced exponential growth over the last decades. Recently, fourth-generation (4G) systems such as IEEE’s worldwide interoperability for microwave access (WiMAX) systems or 3 GPP’s long term evolution (LTE) systems have been deployed. In these systems orthogonal frequency division multiple access (OFDMA) is considered in the physical layer, which divides a whole frequency bandwidth into multiple narrow bandwidths, referred to as subchannels (CHs). As the multiple CHs can be dynamically aggregated for each user, the available resources can be flexibly shared by multiple users to optimize the data throughput.
Another trend on wireless communication is decentralized architectures.
That is, existing cellular networks, i.e., centralized infrastructure, can be extended by supplementary network elements such as femtocells, multihop networks, etc. For example, wireless multihop networks defined in standards such as IEEE’s wireless local area network (WLAN), WiMAX and LTE can be used for building short living network topologies for short term events or for setting up ad hoc topologies in areas where installation of fixed infrastructure is difficult to be deployed. However, due to unreliable ad hoc network topology, it is difficult to reliably provide quality of service (QoS).
In order to overcome the problems in the wireless systems and exploiting their potentials, multimedia types with flexible bit rate are desirable. Currently, 3 GPP recommends to use H.264/AVC [1] for video services such as packet- switched streaming services, multimedia broadcast/multicast services (MBMS) [2], etc. However, the supported video quality is currently quite restricted, and terminals are significantly enhanced in terms of display, processing power, etc.
Hence, a bit rate/quality-scalable extension to H.264/AVC with a backward- compatibility for better access networks, better network conditions and/or high-end receivers is desired.
2
The scalable video coding (SVC) [3] standard as an extension of H.264/AVC allows efficient, standard-based scalability of temporal, spatial, and quality resolution of a decoded video signal through adaptation of the bit stream. Specifically, a single video sequence can be encoded to generate a base and several enhancement layers (BL and ELs). For continuity of video playback, it is enough to receive the BL, whereas the ELs are only used to increase the quality of the rendered video. The ELs can be dropped if the throughput available for streaming the video layers decreases so that stable video service can be provided over the heterogeneous networks.
However, if the system is not adequately manipulated for bit rate requirements of the SVC streams, improvement in service quality cannot be obtained. For example, even if the BL and all the ELs of an SVC stream are transmitted to a receiver for improving the video quality, the received layers cannot be decoded into images if channel error occurs on the BL. In case of multi-user systems, if the system fails to deliver the BLs of the requested SVC streams to some of the users, the overall service quality degrades even if a large number of the ELs are successfully delivered to other users and improve the video quality. Therefore, when transmitting the SVC streams, the provision of the BLs has to be guaranteed for providing the lowest video quality at least.
For example, sufficient transmission time can be allocated for reliable delivery of the BL, and then, remaining time to deadline, i.e., time constraint required for real time service, can be used for transmitting the ELs. In case of multi-user systems, the bandwidth resources can to be fairly allocated to the users for providing the BLs to all the users, and then, the remaining resources can be used for transmitting the ELs, based on channel conditions or requests of the users.
Some of such considerations for various transmission schemes are discussed in this thesis. Specifically, we consider the cases of providing unicast video service (e.g., video telephony, video-on-demand services, clipcasting) and multicast video service (e.g. mobile TV service, interactive video game, live TV for sport events), using the SVC and various wireless systems, i.e., a multihop system, a multi-channel system and a cooperative multi-base station (BS) system. For each case, we consider improving the service quality in terms of received video quality, where the video quality results from received video
Chapter 1. Introduction
3
layers without error. In case of the unicast service, adaptive error control for each layer can be considered based on the priority, i.e., increase in video distortion due to error on the layer. In this manner, the received video quality can be improved, lowering the probability that error occurs on the BL and even the lowest quality video cannot be obtained.
In case of the multicast service, on the other hand, the error probability of the users requesting the same layer can be multiple, and hence, it is difficult to gain from adaptive error control over the layers. Thus, guaranteeing a constant and sufficiently low error criterion for all the transmitted layers, we consider bit rate control by flexibly allocating the bandwidth resources according to the worst user’s channel condition so that each layer transmitted is reliably received by all the targeted users. In this case, therefore, the video quality results only from allocated bit rate (decided by the worst user channel condition) rather than error probability. By adequately allocating the bandwidth resources according to the bit rate requirements of the SVC streams, we can multicast the BL to all the users and the ELs to a part of users based on the channel condition of each user. In this manner, we can improve the total video quality of the users.
Using these concepts of utilizing the SVC for the unicast and the multicast video services, we propose three algorithms for improving the service quality of the three wireless systems. More specific descriptions of the problem statements and the proposed algorithms are given as follows.
The wireless multihop links are connected by wireless links with time varying channel conditions. Therefore, methods for efficiently delivering the video streams, including the SVC streams, by controlling the multihop link parameters such as routing, retransmission concerning path diversity, based on link conditions have been considered. Nevertheless, if we assume a small number of intermediate nodes, we cannot gain from such parameter controls, and therefore, some efficient link adaptation has to be considered.
In Chapter 3, we propose a link adaptation algorithm for further improving expectation of received video quality by using more intelligent reliability criteria according to the relative importance of the layers, where SVC streaming over a communication path configured with a number of wireless links between a source and a destination node is considered. In order to
4
improve the expected video quality, we express the expected video distortion in terms of packet drop rate decided by the source node and packet error rate for each link. Next, we derive an optimization formula for minimizing the expected video distortion subject to time constraints for continuity of video playback. Then, we present an algorithm for minimizing the expected video distortion by adjusting the packet drop rate and the modulation levels of the links. In this manner, adaptive link adaptation can be used so that the received video quality over multihop can be improved.
Next, we consider multicast service such as MBMS, broadcast/multicast service (BCMCS) and multicast broadcast service (MBS), which are respectively included in 3 GPP, 3 GPP 2, and WiMAX standards. In multicast cases, the reception conditions of individual users might be quite different, and therefore, the quality of multicast services suffers from capacity limitation problem, i.e., the multicast bit rate is constrained by the worst-channel user.
However, such multicast modes do not consider efficient dynamic utilization of physical layer resources, and the quality of service (QoS) for multimedia applications, e.g., video services, is in general not sufficiently high to support a large portion of the users.
In Chapter 4, therefore, we address a dynamic resource allocation problem how an OFDMA-based cellular systems, e.g., LTE and WiMAX systems, can maximize average video quality of users requesting SVC streams subject to a total power constraint, where each SVC stream is considered as an MBMS for multiple users. In addition, we also consider the situation where users requesting regular services coexist so that the BS has to guarantee the required bit rates for unicasting to the regular-service users. More specifically, we aim at achieving three objectives. First, for each regular-service user, the system must guarantee a fixed bit rate required for the regular service. Second, for each user that requests an MBMS, the system must guarantee the minimum bit rate required for achieving the minimum video quality. Third, the system must maximize the average video quality of the users requesting the MBMSs. In order to achieve the three objectives, we present a resource allocation algorithm that considers dynamic multicast grouping for each CH in order to mitigate the aforementioned worst user problem, and to achieve the three
Chapter 1. Introduction
5
objectives. In this manner, the satisfaction of the users with heterogeneous channel conditions and service requests can be improved.
In addition, we consider cooperative multi-BS system using OFDMA scheme, referred to as MBMS single frequency network (MBSFN), which is defined in long term evolution (LTE) standard [4]. In an MBSFN, multiple BSs are used to simultaneously multicast the same MBMS signal. As the copies of the same signal are aggregated at the receiving users, the MBSFN transmission enhances the channel quality, and therefore, leads to improvements in spectral efficiency. For reliable and high speed services, application layer FECs, e.g., Reed-Solomon and Raptor codes, are currently used. However, due to the worst user problem, the coverage of MBSFN system is still limited.
In Chapter 5, therefore, we address a resource allocation problem that how an MBSFN can further improve the spectral efficiency subject to a total power constraint. As the multicast bit rate is constrained by the bit rate of the worst- channel user in a multicast group, the worst channel quality for each CH needs to be improved in order to enhance the spectral efficiency. As the worst- channel user and its channel quality for each CH depend on the power balance among the cooperative BSs, we propose a two-phase resource allocation algorithm. First, the algorithm finds the power allocation ratios among the BSs for each CH, which maximizes the worst channel quality for any total power level allowed for that CH. Then, the algorithm distributes the total power over the CHs, where power allocated to each CH again distributed over BSs according to the power allocation ratios for that CH. In this manner, the channel condition of the worst user is enhanced, and service quality can be improved.
The remainder of this thesis is divided as follows. Before we present the three algorithms proposed by this thesis, we review the related researches about resource allocation algorithm for the multi-hop system, the multi- channel system (OFDMA) and the cooperative multi-BS system (MBSFN) in Chapter 2. Then, in Chapter 3, we introduce the first algorithm: modulation level allocation for sequentially transmitting the SVC layers over multi-hop wireless channels. Next, in Chapter 4, we present the second algorithm: power allocation for an OFDMA-based cellular system that serves the regular service
6
and MBMS. Then, in Chapter 5, we present the third algorithm: two-phase power allocation for improving the quality of multicast service provided by an MBSFN. Then, we conclude this thesis in Chapter 6.
Chapter 2. Overview of Resource Allocation Algorithms
7
Chapter 2
Overview of Resource Allocation Algorithms
In this chapter, we give an overview of resource allocation algorithms which have been proposed for efficiently delivering real-time data over multi- hop channels (Section 2.2) and OFDMA channels (Section 2.3, 2.4 and 2.5). In order to deliver video data over time varying wireless channels, SVC, which provides flexible bit rate for a video content, is considered in many algorithms for wireless video transmission. In Section 2.1, we review the bit rate adaptability of the SVC stream. Then, from Section 2.2 to Section 2.5, we review existing algorithms related with algorithms proposed in this thesis.
Thereafter, in Section 2.6, we explicit the difference of the proposed algorithms from the previous algorithms.
2.1. Scalable Video Coding
The SVC is designed to provide adaptation in required bit rate and decoded video quality1 . That is, a sequence of video images is compressed into multiple bit streams referred to as video layers each of which is a replica of the sequence with different compression ratio from other layers, resulting in different bit rate and video quality. Hence, the bit rate can be adjusted by dropping some of the layers, gracefully degrading the received video quality.
More specifically, the video layers that represent basic information and details are referred to as base layer (BL) and enhancement layer (EL), respectively.
Figure 2.1 shows an example of hierarchical-B coding structure [3] of a group of pictures (GOP) configured with nine frames and three quality layers (QLs).
1 In this thesis, we use peak signal to noise ratio (PSNR) for measuring the video quality [1]
8
Figure 2.1 Coding structure of a GOP configured with nine frames and three quality layers, where the two indicators in each box denote the frame and the quality layer
indices, respectively.
Each box in Figure 2.1 denotes an SVC packet, where the two numbers in each box denote the frame (image) and the QL indices. Each video frame is coded into BL packet with QL index 1 by using the highest compression ratio, i.e., the lowest bit rate required and video quality. Among the frames, temporal prediction of the hierarchical-B structure can be considered in order to improve the coding efficiency and to enable the temporal scalability. For each frame, information for improving the video quality is coded into EL packets with QL index larger than 1. As the BL and the EL packets of each frame are coded in progressive manner, each packet cannot be decoded without the packets with lower QL index. By using the three QLs, we can obtain three PSNR options.
We obtain the lowest PSNR by decoding the BL and improve the PSNR by additively decoding ELs in ascending order of the QL index. This quality scalability scheme is referred to as coarse grain scalability (CGS). The QLs are also referred to as CGS layer in the rest of this thesis.
The standard SVC allows decoding each packet of EL independently from other packets in EL in order to achieve more bit rate options. This quality scalability scheme that allows packet-based rate control is referred to as medium grain scalability (MGS). In the rests of this thesis, the packets in ELs are referred to as MGS packets. In order to achieve more efficient bit rate adaptabilities in terms of bit rate-PSNR performance, transmission priorities can be decided for the MGS packets according to the bit length and the
1, 1 2, 1 3, 1 4, 1 5, 1 6, 1 7, 1 8, 1 9, 1
1, 2 2, 2 3, 2 4, 2 5, 2 6, 2 7, 2 8, 2 9, 2
1, 3 2, 3 3, 3 4, 3 5, 3 6, 3 7, 3 8, 3 9, 3
Temporal prediction (Hierarchical-B structure) Inter quality layer prediction
Chapter 2. Overview of Resource Allocation Algorithms
9
contribution to the PSNR of each MGS packet [5]. For example, nineteen bit rate options can be achieved by the coding structure in Figure 2.1. Figure 2.2 shows an example of the bit rate options coded by joint scalable video model (JSVM) [6], where first nine frames of test videos “Football” and “Crew”, common intermediate format (CIF) and 30 frames per second (fps), are used.
The quality layers are coded with video quantization parameters (QPs) 40, 34 and 28, respectively, where higher compression ratio can be obtained by higher QP.
Figure 2.2 Bit rate-PSNR performance test videos “Football” and
“Crew” (CIF, 30 fps) coded by JSVM.
For each test video, three performance points of CGS and additional sixteen points by considering MGS can be achieved. As shown in Figure 2.2, we can obtain the bit rate-PSNR flexibility by considering SVC for transmitting over time-varying channel conditions.
2.2. Video Transmission over Multi-hop Channels
In this section, we review algorithms proposed to transmit video data, including SVC stream, over multi-hop channels which are considered for expanding coverage of wireless services. In [7], the authors analyzed behaviors of queue delay over the paths between a couple of source and destination nodes.
If the SVC packets are received by the destination node after intended deadline,
28 30 32 34 36 38 40
0 5 10 15 20 25 30 35
Footba ll Crew
PSNR [dB]
Bit rate [10 Kbps]
CGS Points
10
the packets are discarded and the video quality at the destination node degrades.
Therefore, the routing algorithm for delivering the packets in time is considered by forwarding the packets according to the remaining time of each packet. Since wired networks are considered in [8], they assume that link failure does not occur, and therefore, the delay over deadline is the only factor resulting in the received video quality.
In [9-11], SVC packets are assumed to be transmitted over networks comprising wireless links with various link failure probabilities due to unreliable characteristics of wireless channels. Given the link failure probabilities and deadline, the routing algorithms for minimizing expected video qualities at the destination node are considered. In these papers, retransmission is also considered for compensating the link failures, where transmitting replicas over multiple links is also considered to reduce the probability of link failures.
In [12], broadcasting characteristic of wireless channels is utilized by considering opportunistic routing. That is, when a node on a path transmits a packet, all nodes closer to destination node than the transmitting node hear the transmitted signal. Then, the closest node to the destination node among nodes which successfully received the packet can forward the packets. The performance of the opportunistic routing is calculated by modeling the patterns of the nodes successfully receiving the packet by Markov chain on the basis of the link failure property of each link.
Video compression ratio also affects the received video quality over multi- hop networks [13]. Therefore, video compression ratio also can be adapted for improving the received video quality over multi-hop networks as discussed in [14]. In [14], the authors assume a video packet can be divided and separately streamed over multiple paths, where the total size of the video session can be controlled by the compression ratio. The coding ratio and the size of video flow over each path is determined to minimize the expected video distortion, which is the sum of three types of video distortion, induced by video coding, link failure and delay over deadline.
In the standard video coding [1], slice coding is provided, which divides an image (also called frame) into a number of spatial segments (called slices) as shown in Figure 2.3(a) and codes each segment separately. The quality of
Chapter 2. Overview of Resource Allocation Algorithms
11
each image at the destination node results from the successfully received slices.
In [15], the slice coding is considered for obtaining path diversity gain by separately transmitting each slice over separate paths, where the size of each slice also can be adapted by controlling the compression ratio. However, if a slice is not received, only the spatial part that the lost slice represents degrades, and perceptual quality can be degraded severely.
Figure 2.3 Examples of (a) slice coding and (b) multiple description coding.
Therefore, multiple description coding (MDC) [16, 17] has been considered for video transmission systems. The MDC divides an image into multiple images called descriptions, where each description describes the whole image with degraded quality. For example, an image can be divided into two descriptions, one with odd pixels and the other with even pixels as shown in Figure 2.3(b). If only one of the descriptions is received, the remaining pixels can be filled by interpolating the pixels in the description. The main difference of the MDC from the SVC is that there are not priorities among the descriptions, and therefore any received MDC packet can be decoded independently from other MDC packets, whereas, in case of the SVC, whether one packet can be decoded is dependent on other packets. In [18-20], two descriptions are considered for transmitting each video session over two paths between source and destination nodes. Therefore, routing algorithm for the two paths is considered for minimizing the expected video quality. In this case, there are four patterns of the descriptions received at the destination, which means four video qualities. On the basis of these video qualities, the expected video quality can be quantified according to the link failure and the delay over deadline.
Description 1 Description 2
(b) Slice 1
Slice 2
Slice 3 (a)
12
In this section, we overviewed the previous algorithms for transmitting video data over multi-hop networks, which are mostly considered for finding optimal path(s), where packet error rate and affordable bit rate of each link are given. However, if there is only one feasible path, we cannot benefit from such algorithms. In this case, link adaptation can be considered. For example, in [21-22], link adaptation for transmitting SVC streams over a single link is considered. These link adaptations and the proposed link adaptation for multihop channel are discussed in Section 2.6.1.
2.3. Resource Allocation for OFDMA
OFDM is one of the promising solutions for high speed transmission and has been widely adopted in wireless standards. In wireless signal transmission, multipath fading is a common phenomenon especially in urban areas where the communication environment changes quickly. In those areas, multipath propagations occur from different objects which results in the electromagnetic wave travelling along different paths of varying length. The interaction between those waves causes multipath fading with frequency selectivity where the fading parameter changes with frequency. As a result, the wireless channel is assumed to be time-varying frequency-selective multipath fading.
An extensive overview of statistical analysis and information-theoretic and communications features of fading channels has been presented in [23]. The multipath fading channels in frequency domain can be characterized by coherence bandwidth of the channel defined as a range of frequencies over which the channel can be considered flat [24]. That is, by choosing the bandwidth of the subchannels (CHs) of OFDM much smaller than the coherence bandwidth, each CH can be assumed to undergo flat fading.
Figure 2.4 Quality of OFDMA CHs over frequency selective channel.
1 2 3 4 5 6 7 … CH
Channel quality
1 2 3 4 5 6 7 … CH
Channel quality
User 1 User 2
Chapter 2. Overview of Resource Allocation Algorithms
13
Therefore, many algorithms for power and rate allocation over CHs have been proposed for improving the service quality of OFDMA channel. In the remainder of this section, we first review the resource allocation algorithms for unicast service (i.e., each transmission is intended for one user at most). Then, existing resource algorithms for multicast service (i.e., each transmitted signal can be received by multiple users) on the basis of the resource allocation algorithms for the unicast service are discussed.
In a single user system, the user can use the total power to transmit on all CHs. The system is then optimized by exploiting the frequency selectivity of the channel and dynamically adapting the modulation type and transmits power on each CH. These dynamic power allocation schemes [25-26] have shown significant performance gain in terms of throughput compared to the case that the total power is equally distributed over all the CHs. The OFDM channel also can be utilized for multiple-access channel by allocating the CHs to different users. In [27], it is shown that the bit rate of a multiuser OFDM system is maximized when each CH is assigned to the user that has the best channel gain for that CH and the total power is distributed among the CHs according to water-filling [28].
(2.1)
(2.2)
where and are bit rate for user and power allocated to CH . However, when we consider multiuser systems, not only efficiency but fairness is also crucial issue in resource allocation. Fairness could be defined in terms of different parameters of the system. The fairness can be considered in terms of bandwidth, i.e., each user is assigned an equal number of CHs [29-30].
It also can be considered in terms of bit rate, where the objective is to allocate the resources to the users such that all the users achieve the same bit rate [31].
Also, different types of constraints for fairness can be considered. For example, proportional bit rate constraints as stated in (2.3) are considered in [32-34], whereas the minimum bit rate constraint for each user as stated in (2.4) is considered in [35], in addition to the power constraint in (2.2), where the terms
14
and are bit rate proportion and minimum bit rate constraint for user .
(2.3)
(2.4)
In [36], bandwidth constraint for each user is assumed by constraining the maximum frequency gap between two CHs of the highest and the lowest frequency. Different types of rate constraints can be considered for multiple users. For example, in [37], coexistence of real time and non real time users are considered, where the minimum bit rate for each of the real time users has to be guaranteed. The rest resource can be utilized for maximizing the total bit rate of the non real-time users. While the above system objectives are maximizing the bit rate, in [38-43], total power minimization is considered, where minimum bit rate for each user is considered.
(2.5)
guaranteeing the minimum bit rate for each user as stated in (2.4).
In [44-47], concept of utility functions, denoted as , is considered to formulate the problem of resource allocation in multiuser OFDM systems.
Utility function can be defined according to system objectives (e.g., can be video quality of SVC stream, achieved by bit rate ). The utility-based dynamic resource allocation problem is formulated as
(2.6)
If the utility is monotonically increasing concave function of bit rate, we consider convex optimization. For example, a logarithmic utility function is considered in [48].
Chapter 2. Overview of Resource Allocation Algorithms
15
In [49-50], video transmission over OFDMA is considered by using the weighted sum maximization algorithm, where the bit rate allocation ratio among multiple users can be adjusted by controlling the weights. More specifically, the weights for transmitting SVC packets of multiple users are controlled according to remaining time of each packet until deadline and increase in video quality by transmitting each packet. The OFDMA resource allocations for video transmission are also considered in [51-52]. In [52], the OFDM resource allocation for minimizing the total video quality is considered.
In [51], two resource allocation algorithms for transmitting SVC over the OFDMA channel is considered, which are focused on fairness and efficiency in terms of video quality. By operating the two algorithms together, the fairness and efficiency performances can be jointly obtained.
2.4. Resource Allocation for Multicast Service
As mobile devices developed, user demands for multimedia applications, such as mobile TV or video game, have been rapidly grown. Accordingly, multicast services, i.e., MBMS, over cellular systems have been considered and standardized [4]. In [53-59], MBMS over universal mobile telecommunications system (UMTS) is considered. In these researches, the channel structures for multicast, performance improvement by considering adaptive MCS, and user grouping for multicast users are analyzed.
In [60], OFDMA channel is considered for MBMS. In this research, tradeoff between system throughput maximization and error probability according to MCS is analyzed. For a given MCS, the error probability can be compensated by improved transmit power level. Therefore, resource allocation algorithm for OFDMA can be used for multicasting, where the bit rate of multicast group over CH , denoted as , can be determined by
(2.7)
where the terms and denote indices of users in multicast group and signal to noise ratio (SNR) of user over CH , respectively. For
16
example, in [61], modulation and coding scheme (MCS) and power allocation algorithm for multicasting over OFDMA channel is considered. The objective of the algorithm in [61] is to maximize the weighted sum of bit rates for the users. In [62-63], video services using the SVC are considered, where the minimum bit rates for transmitting the BLs, the essential parts of the SVC streams (see Figure 2.1), has to be guaranteed aiming all the users. Algorithms are designed to find resource allocation that maximizes the total bit rate while guaranteeing the minimum bit rates.
2.5. Resource Allocation for MBSFN
There have been performance evaluation and resource allocation algorithms for the MBSFNs. In [64], OFDM cells in the MBSFN mode and normal mode are considered, where the cells with the normal modes access the OFDM channel in time division manner. In this research, the authors found that if the MBSFN zone is expanded by using more cells for MBSFN, average bit rate is improved. In this case, however, the flexibility in transmitting the data is degraded, because the cells in an MBSFN zone are dedicated to the same set of service. In [65-66], the authors considered configuring MBSFNs (i.e., how to group the BSs into MBSFNs) such that the number of the time- frequency resource blocks of OFDM channels can be minimized. While a fixed MCS is considered for all the resource blocks in [65-66], CH and MCS allocation for the multiple MBSFNs are considered in [67], where equal power allocation are assumed. Adaptive power allocation is also considered for the cooperative BSs for unicast [68] and multicast [69] service. In [69], as the copies of the same signal are aggregated at the receiving users and the bit rate is mainly decided by the worst-channel user, the bit rate of multicast group can be determined by
(2.8)
where is SNR for user from BS over CH .
Chapter 2. Overview of Resource Allocation Algorithms
17
2.6. Proposed Resource Allocation Algorithms
In this section, we discuss the three proposed-algorithms, comparing with the previous algorithms. We first discuss the first algorithm which is considered to transmit SVC stream over a path comprising multiple wireless links between source and destination nodes in Section 2.6.1. The objective of this algorithm is to minimize expected video distortion at the destination node.
In Section 2.6.2, we discuss the second algorithm which is considered to provide multicast and unicast service over an OFDMA-based BS. The objective of this algorithm is to provide a fixed required bit rate for uinicast service and maximize total video quality for multicast service, where we assume the multicast service is for streaming SVC packets. In Section 2.6.3, we discuss the last algorithm which is considered for multicast service over MBSFN. The objective of this algorithm is to maximize system utility, where utility of each MBMS can be chosen according to the system objective.
2.6.1. Proposed Algorithm in Chapter 3
In Section 2.2, we overviewed the previous algorithms for transmitting video data over multi-hop networks, which are mostly considered for finding optimal path(s) to minimize the video distortion at the destination node. In these researches, packet error rate and affordable bit rate of each link, which affects on deciding the optimal path(s), are given by the channel quality and MCS for the link. When there are a sufficient number of nodes, and therefore, multiple paths exist, these algorithms can be utilized for improving the performance of the video transmission. However, if there is only one feasible path between source and destination node, we cannot benefit from such algorithms.
With the algorithm proposed in Chapter 3, we focus on the case that there is only one feasible path between source and destination node, where modulation level for each link can be adjusted for improving the video quality at the destination node. If the SVC is considered for this scenario, the bit rate required for transmitting the SVC stream can be adjusted according to the modulation levels of the links on the path, by adequately dropping the MGS
18
packets (See Figure 2.1). For example, if the channel quality of the links is high, then MCSs with higher spectral efficiency can be considered, maintaining packet error probability low, and hence, more MGS packets can be transmitted for improving the video quality at the destination node. However, if the channel quality is not high enough for transmitting the SVC packets, and, unfortunately, link error occurs on a packet with high priority, e.g., packet (1, 1) in Figure 2.1, then all the received SVC packets cannot be decoded and video quality at the destination node severely degrades. In this case, we need to consider more reliable modulation level with lower spectral efficiency for preventing the packets with higher priority from packet error, even if more MGS packets have to be dropped and the received video quality can be degraded.
This tradeoff between packet error rate and the packet dropping control is studied in [21-22] for a single link. In these algorithms, packet error rate adapted to the packet priority is considered. More specifically, the authors consider allocating modulation levels with lower spectral efficiency to the packets with higher priority in order to prevent the server damage in video quality, while allocating modulation levels with higher spectral efficiency to the packets with lower priority for transmitting more packets in deadline to improve the video quality, where packet error on the lower priority packets does not result in severe quality degradation. In this manner, they aim to minimize expected video distortion at the receiver side.
On the basis of the algorithm of [22], we consider minimizing the expected video distortion of the SVC stream transmitted over a path consists of multiple links by determining optimal modulation level of each link and packet dropping control at the source node. The expected distortion can be quantified by the packet drop rate and the modulation levels of the links for each packet.
Then, the expected distortion is minimized by the proposed algorithm presented in Chapter 3.
2.6.2. Proposed Algorithm in Chapter 4
In Section 2.5, OFDMA resource allocation for multicast service is overviewed, which can be formulated on the basis of the resource allocation
Chapter 2. Overview of Resource Allocation Algorithms
19
problem in Section 2.4 for unicast service. In [62-63], guaranteeing the minimum bit rates for transmitting the BLs, i.e., the essential parts of the SVC streams, to all the users is considered. More specifically, the algorithms are designed to find resource allocation that maximizes the total bit rate while guaranteeing the minimum bit rates, where the total power is constrained below a predefined power level. This previous algorithms split the CHs into two sets, a set to guarantee the minimum bit rates and the other set to maximize the total bit rate.
Although, these algorithms provide efficient resource allocation, avoiding undesirable cases that any user is not available of the BL as far as possible, the given resources still can be wasted, because maximizing the total bit rate is not directly related to maximize users’ satisfaction, i.e., PSNR. For example, a large part of the resources can be used to allocate the bit rates exceeding the maximum bit rates (required for transmitting all the video packets) to some of the users, while allocating only the minimum bit rates to other users. In addition, the multicast configuration can have the users difficult to decode the received bits properly, where the detailed explanation is provided in Section 4.1.2.
In Chapter 4, we consider maximizing total PSNR rather than total bit rate for multicasting SVC stream, guaranteeing the minimum bit rate for transmitting the BLs for all the users. In addition, we consider more realistic multicast configurations for multicasting the SVC streams. The detailed presentation is provided in Chapter 4.
2.6.3. Proposed Algorithm in Chapter 5
As reviewed in Section 2.6, OFDMA resource allocation over an MBSFN is proposed in [69]. The objective of this algorithm is to maximize the bit rate, where total power consumed by each BS is constrained as
(2.9)
20
where is power from BS over CH . As discussed in [69], the optimal power of each BS over each CH for maximizing the bit rate is dependent on the power of the other cooperating BSs over the same CH. Hence, the authors considered iteratively updating the power allocation of each BS in turn until the resource allocation become stable. In this manner, total power given to each BS can be distributed so that the bit rate performance can be improved.
In order to improve the system performance further, we also consider power distribution among the BSs by assuming more flexible power constraint for the cooperative BSs serving an MBSFN. More specifically, we consider a single constraint on the power consumed by all the BSs as (2.10) so that larger portion of the power can be allocated to BSs with higher user density around.
(2.10)
As multicast transmission rate is constrained by the transmission rate of the worst-channel user in the multicast group, the worst channel quality for each CH needs to be improved in order to improve the spectral efficiency. In Chapter 5, we will show that the worst-channel user and its channel quality for each CH are depending on the power balance among the cooperative BSs.
Therefore, in Chapter 5, we propose a two-phase resource allocation algorithm.
First, the algorithm finds the power allocation ratios among the BSs for each CH, which maximizes the worst channel quality for any total power level allowed for that CH. Then, the algorithm distributes the total power over the CHs, where power allocated to each CH again distributed over BSs according to the power allocation ratios for that CH.
Chapter 3. Modulation Level Allocation for Unicast Service over Multi-hop Channels
21
Chapter 3
Modulation Level Allocation for Unicast Service over Multi-hop Channels
In this Chapter, the time resource allocation method for efficiently transmitting MGS video packets over a transmission path consisting of multiple wireless links is introduced. As explained in Chapter 2, MGS provides bit rate adaptation according to the available bit rate by dropping a number of video packets in the compressed bit stream. In other words, rate- distortion control can be achieved by means of packet transmission control.
The available bit rate and the spectral efficiency are determined by the bandwidth and the modulation level, respectively. Accordingly, the number of packets available for transmission is affected by the modulation level of the packets. However, if we consider modulation levels with higher spectral efficiency in order to increase the number of packets and reduce the expected video distortion, the packet error rate of the transmitted packets can also be increased because the spectrally efficient modulation levels are sensitive to channel noise. This is another reason for the increment in expected video distortion, because the erroneous received packets cannot be used for video reconstruction. Therefore, we consider the minimization of expected video distortion via the optimization of two factors: packet extraction for transmission, and modulation level allocation for the extracted packets. Packet extraction is optimized for the path between the source and destination nodes, whereas the modulation level for each extracted packet is optimized for each link along the transmission path.
3.1. Expected Video Distortion
Video distortion of decoded frames in a group of pictures GOP is affected by the combination of packets available for the video reconstruction. Therefore, it is necessary to predict the combination at the destination node in order to
22
control and reduce the distortion. The combination of the received packets at the destination node for each packet in a GOP results from two factors: the packet drop rate (PDR) (decided by the transmitter) and the PER (influenced by channel noise). We define the packet loss rate (PLR) as the probability that the packet is not available at the destination node, given by where and are the PER and PDR, respectively. For packets in a GOP, the number of combinations is , as two cases (that of being used and unused for decoding) can be considered for each packet.
Therefore, the expected distortion of the th frame is
(3.1)
where and are the probability that the th combination occurs at the destination node and the distortion of the th frame in the th combination, respectively. Equation (3.1) implies that decoding simulations are required to calculate , because must be measured for by the decoding simulations. can be expressed in terms of the PLR of packets ( for ) as
(3.2)
where denotes whether is to be used ( = 1) or not ( = 0).
3.1.1. Simple Examples of Expected Distortion
Let us assume that there are three frames, each of which is coded to one packet, as depicted in Figure 3.1. In this figure, the packet reference [3] is expressed by the arrows.
