10
Bit Allocation for Scalable Video Streaming over Mobile Wireless Internet Fan Yang 1 *, Qian Zhang 2 , Wenwu Zhu 2 , Ya-Qin Zhang 2 1 Dept. of Computer Science, Nanjing University,No. 22, Hankou Road, Nanjing, 210093, China 2 Microsoft Research Asia, No. 49, Zhichun Road, Haidian District, Beijing, 100080, China email: [email protected], {qianz,wwzhu, yzhang}@microsoft.com Abstract—With the convergence of wired line Internet and mobile wireless networks as well as tremendous demand on video application in mobile wireless Internet, it’s essential to design an effective video streaming protocol and an efficient resource allocation scheme for video delivery over wireless Internet. In this paper, we employ WMSTFP, an end-to-end TCP-friendly multimedia streaming protocol, to detect the status of the wired and wireless part of the wireless Internet, where only the last hop is wireless link. By accurately distinguishing the packet losses due to transmission errors from the congestive losses and smoothing out the pathologic round-trip-time values caused by the highly dynamic wireless environment, in WMSTFP higher throughput in wireless Internet can be achieved and rate can be adjusted in a smooth and TCP-friendly manner. Based upon WMSTFP, we propose a novel loss pattern differentiated bit allocation scheme while applying unequal loss protection (ULP) for scalable video streaming over wireless Internet. Specifically, a Rate-Distortion (R-D) based bit allocation scheme which considers both wired and wireless network status is proposed to minimize the expected end-to-end distortion. The optimal solution of global optimization for the bit allocation scheme is obtained by a local search algorithm taking the characteristics of Progressive Fine Granularity Scalable (PFGS) video into account. Analytical and simulation results demonstrate the effectiveness of our proposed schemes. I. INTRODUCTION Streaming video over Internet has been very active for the past several years [1] and has been commercialized. Recent advances in wireless communication technology and the increasing computing capability of mobile devices make streaming video over wireless of great interest [2]. As different types of wireless networks are converged into all IP networks and into the Internet, it’s important to study video streaming over the converged wireless Internet. It is well known that bit error rate (BER) of wireless network is much higher than that in wired-line links. Moreover, the varying wireless environment also results in dramatic fluctuation of network bandwidth and delay. Specifically, the bandwidth and delay may vary with the changing distance between the base station (or access point) and mobile host. Furthermore, link layer error control scheme like automatic repeat request (ARQ) is widely used to overcome the wireless * Work performed when Fan Yang was a visiting student at Microsoft Research Asia. channel errors. This will further increase the dramatic variation of bandwidth and delay in wireless network. In mobile wireless Internet, all the above characteristics bring great challenges for video streaming. We argue that video streaming applications should be aware and adaptive to the change of network conditions of the wireless Internet, i.e., quality of service (QoS) fluctuations, so as to get good video quality. This adaptation consists of network adaptation and media adaptation. Network adaptation refers to how many network resources (e.g., bandwidth) a video streaming application should utilize for video content, resulting in designing an adaptive streaming protocol for video transmission. Usually a media streaming protocol chooses its transmission speed in accordance with the measured packet loss ratio and round trip time (RTT). Therefore, the correct network adaptation relies on the accurate measurement of packet loss ratio and round-trip-time. However, it’s not easy to accurately measure the two parameters in wireless Internet as it is in wired-line networks. Firstly, in wireless Internet scenario, the packet loss, which is considered to be a signal of network congestion, can be caused by either network congestion occurred in wired networks or transmission errors occurred in wireless links. Although we can install an agent at the edge of wired and wireless network to discriminate congestive losses from erroneous losses [3-7], the additional intermediate agents introduce the deployment issue for network operators. We could also adopt some heuristic scheme such as inter-arrival time and packet pair to differentiate packet loss type from end hosts [8-11]. But the accuracy of the loss type differentiation is doubtful. Secondly, in order to minimize the reverse traffic, many streaming protocols acknowledge only one packet to measure the RTT within a pre-determined period of time. However in wireless environment, the aforementioned bandwidth and delay fluctuation cause a dramatic variation of RTT value. That is to say, the rate estimation based upon RTTs may not be able to reflect the correct status of the wireless networks and could suffer from great jitters. Another important issue is that the designed streaming protocol should be friendly to TCP protocol that is widely used in today’s Internet [12, 13]. We employ WMSTFP, the end-to-end TCP-friendly multimedia streaming protocol for a network where only the last hop is wireless link, to address the above issues. WMSTFP can differentiate the loss type in an end-to-end manner based upon the link layer information and can characterize the packet loss behaviors taking the loss burstiness into account. Moreover, it can filter out the abnormal RTT values caused by the highly varying wireless environment. Note that a 0-7803-8356-7/04/$20.00 (C) 2004 IEEE IEEE INFOCOM 2004

Bit Allocation for Scalable Video Streaming over Mobile

  • Upload
    ronny72

  • View
    736

  • Download
    0

Embed Size (px)

DESCRIPTION

 

Citation preview

Page 1: Bit Allocation for Scalable Video Streaming over Mobile

Bit Allocation for Scalable Video Streaming over Mobile Wireless Internet

Fan Yang1*, Qian Zhang2, Wenwu Zhu2, Ya-Qin Zhang2 1 Dept. of Computer Science, Nanjing University,No. 22, Hankou Road, Nanjing, 210093, China

2 Microsoft Research Asia, No. 49, Zhichun Road, Haidian District, Beijing, 100080, China email: [email protected], {qianz,wwzhu, yzhang}@microsoft.com

Abstract—With the convergence of wired line Internet and mobile wireless networks as well as tremendous demand on video application in mobile wireless Internet, it’s essential to design an effective video streaming protocol and an efficient resource allocation scheme for video delivery over wireless Internet. In this paper, we employ WMSTFP, an end-to-end TCP-friendly multimedia streaming protocol, to detect the status of the wired and wireless part of the wireless Internet, where only the last hop is wireless link. By accurately distinguishing the packet losses due to transmission errors from the congestive losses and smoothing out the pathologic round-trip-time values caused by the highly dynamic wireless environment, in WMSTFP higher throughput in wireless Internet can be achieved and rate can be adjusted in a smooth and TCP-friendly manner. Based upon WMSTFP, we propose a novel loss pattern differentiated bit allocation scheme while applying unequal loss protection (ULP) for scalable video streaming over wireless Internet. Specifically, a Rate-Distortion (R-D) based bit allocation scheme which considers both wired and wireless network status is proposed to minimize the expected end-to-end distortion. The optimal solution of global optimization for the bit allocation scheme is obtained by a local search algorithm taking the characteristics of Progressive Fine Granularity Scalable (PFGS) video into account. Analytical and simulation results demonstrate the effectiveness of our proposed schemes.

I. INTRODUCTION Streaming video over Internet has been very active for the

past several years [1] and has been commercialized. Recent advances in wireless communication technology and the increasing computing capability of mobile devices make streaming video over wireless of great interest [2]. As different types of wireless networks are converged into all IP networks and into the Internet, it’s important to study video streaming over the converged wireless Internet.

It is well known that bit error rate (BER) of wireless network is much higher than that in wired-line links. Moreover, the varying wireless environment also results in dramatic fluctuation of network bandwidth and delay. Specifically, the bandwidth and delay may vary with the changing distance between the base station (or access point) and mobile host. Furthermore, link layer error control scheme like automatic repeat request (ARQ) is widely used to overcome the wireless

* Work performed when Fan Yang was a visiting student at Microsoft Research Asia.

channel errors. This will further increase the dramatic variation of bandwidth and delay in wireless network. In mobile wireless Internet, all the above characteristics bring great challenges for video streaming. We argue that video streaming applications should be aware and adaptive to the change of network conditions of the wireless Internet, i.e., quality of service (QoS) fluctuations, so as to get good video quality. This adaptation consists of network adaptation and media adaptation.

Network adaptation refers to how many network resources (e.g., bandwidth) a video streaming application should utilize for video content, resulting in designing an adaptive streaming protocol for video transmission. Usually a media streaming protocol chooses its transmission speed in accordance with the measured packet loss ratio and round trip time (RTT). Therefore, the correct network adaptation relies on the accurate measurement of packet loss ratio and round-trip-time. However, it’s not easy to accurately measure the two parameters in wireless Internet as it is in wired-line networks. Firstly, in wireless Internet scenario, the packet loss, which is considered to be a signal of network congestion, can be caused by either network congestion occurred in wired networks or transmission errors occurred in wireless links. Although we can install an agent at the edge of wired and wireless network to discriminate congestive losses from erroneous losses [3-7], the additional intermediate agents introduce the deployment issue for network operators. We could also adopt some heuristic scheme such as inter-arrival time and packet pair to differentiate packet loss type from end hosts [8-11]. But the accuracy of the loss type differentiation is doubtful. Secondly, in order to minimize the reverse traffic, many streaming protocols acknowledge only one packet to measure the RTT within a pre-determined period of time. However in wireless environment, the aforementioned bandwidth and delay fluctuation cause a dramatic variation of RTT value. That is to say, the rate estimation based upon RTTs may not be able to reflect the correct status of the wireless networks and could suffer from great jitters. Another important issue is that the designed streaming protocol should be friendly to TCP protocol that is widely used in today’s Internet [12, 13].

We employ WMSTFP, the end-to-end TCP-friendly multimedia streaming protocol for a network where only the last hop is wireless link, to address the above issues. WMSTFP can differentiate the loss type in an end-to-end manner based upon the link layer information and can characterize the packet loss behaviors taking the loss burstiness into account. Moreover, it can filter out the abnormal RTT values caused by the highly varying wireless environment. Note that a

0-7803-8356-7/04/$20.00 (C) 2004 IEEE IEEE INFOCOM 2004

Page 2: Bit Allocation for Scalable Video Streaming over Mobile

InternetStreaming

Client

StreamingServer

Base Station(Access Point)

WirelessAccess

description of WMSTFP and its preliminary simulation results first appeared in [14], we regard it in this paper as a building block in the architecture for scalable video streaming over wireless Internet and practically evaluate its effectiveness using scalable video codec.

Media adaptation is about how to control bit-rate of a video stream and adjust error control behaviors according to the varying wireless Internet conditions. Forward Error Correction (FEC) and ARQ are two well-studied error control techniques. Of these two error control mechanisms, FEC has been commonly suggested for real-time applications due to strict time constraints and the error-tolerance nature of media streams. Studying how to add FEC to scalable video coding is of great interest recently because of scalable video’s scalability in bitstream and therein friendliness to network [15]. A scalable encoder usually encodes a raw video sequence into base layer, which has more important data, and multiple enhancement layers. Because different part of the compressed video bitstream has different importance, naturally for scalable video, we can use FEC codes to protect them with different strength, i.e., unequal error protection. However, to decide how many FEC codes we should allocate to different part of video bitstream is a challenging task in wireless Internet environment. The recent work on unequal error protection and bit-allocation for video delivery over wireless network and wired network can be found in [16-18] and [19-21], respectively. But in their bit allocation schemes, they treat the erroneous losses and the congestive losses the same and only consider one type of packet loss. As stated above, in wireless Internet environment the packet losses consist of both congestive losses and erroneous losses, which in turn have different loss patterns in wireless and wired network parts. As different loss patterns lead to different perceived QoS at application level [22], the aforementioned bit allocation schemes can not achieve good video quality since the joint effects of their loss patterns have not be accurately reflected.

To address those issues in media adaptation, we propose a loss differentiated rate-distortion (R-D) based bit allocation scheme for Progressive Fine Granularity Scalable (PFGS) video [23] streaming over wireless Internet while employing a network adaptive unequal loss protection (ULP) scheme. More specifically, we notice that the packet losses due to network congestion and those caused by wireless errors result in different perceived QoS in video streaming. Consequently, our proposed bit allocation scheme distributes the available bit resource between source bits and channel protection bits by taking the loss pattern difference between congestive packet losses and erroneous packet losses, differentiated by WMSFTP, into account so as to improve the end-to-end video quality.

The rest of this paper is organized as follows. In Section II, we propose an end-to-end architecture for scalable video streaming over wireless Internet. We present a TCP-friendly streaming protocol for video over wireless Internet in Section III. In Section IV, network adaptive unequal packet loss protection and rate-distortion based bit allocation are discussed. Section V gives simulation results and Section V concludes this paper.

II. ARCHITECTURE FOR END-TO-END SCALABLE VIDEO STREAMING OVER WIRELESS INTERNET Figure 1 depicts a general scenario for video streaming over

wireless Internet. In this scenario, streaming server resides in the Internet and streaming client accesses to the Internet through the last mile wireless connection. Since wireless access and Internet backbone have their individual network condition, usually it is difficult to achieve the optimal end-to-end perceived media quality. In this work, streaming protocol and resource allocation are studied in the sender and receiver to achieve such a goal.

Figure 1. General illustration of video streaming over wireless Internet.

Figure 2 depicts the detailed diagram of our system architecture for end-to-end scalable video streaming over wireless Internet. The key components in this architecture consist of WMSTFP Congestion Control and WMSTFP Network Monitor performed by our designed media streaming protocol WMSTFP, Network-adaptive ULP Channel Encoder, and Loss Differentiated R-D Based Bit Allocation. Here we adopt PFGS as an example of layered scalable video.

Figure 2. System architecture for scalable video streaming over wireless Internet.

WMSTFP Congestion Control and WMSTFP Network Monitor provide network adaptation at end hosts, which mainly deal with probing and estimating the dynamic network conditions using WMSTFP. Specifically, the WMSTFP Congestion Control module adjusts sending rate on the sender side based on the feedback information, and the WMSTFP Network Monitor module on the receiver side analyzes the erroneous loss and congestive loss rate caused in wireless connection and wired connection and estimates the end-to-end available network bandwidth. The control data consisting of the estimated network bandwidth and other related network status parameters, such as congestive packet loss rate and erroneous packet loss rate, and smoothed packet transmission time, are fed back to the sender.

Network-adaptive ULP Channel Encoder module protects different layers of PFGS against congestive packet losses and erroneous losses according to their importance and network

Network-adaptiveULP Channel Encoder

WMSTFPCongestion Control

WMSTFP NetworkMointor

Data Flow

Control Flow

Internet

PFGS SourceEncoder

Loss DifferentiatedR-D Based Bit

Allocation Base Station

ChannelDecoder

PFGS SourceDecoderWMSTFP

0-7803-8356-7/04/$20.00 (C) 2004 IEEE IEEE INFOCOM 2004

Page 3: Bit Allocation for Scalable Video Streaming over Mobile

status using Reed-Solomon codes. This is because different parts of PFGS video bitstream have different quality impact, thus it’s desirable to add different protection level according to the importance of the scalable video bitstream.

Loss Differentiated R-D Based Bit Allocation module performs media adaptation control to make the total sending rate adapt to the estimated network conditions. Specifically, based on the feedback information from the receiver, this module on the sender side distributes the total sending rate between video bit rate and error protection rate according to the available bandwidth and different packet loss conditions in wired and wireless connections.

III. MULTIMEDIA STREAMING TCP-FRIENDLY PROTOCOL FOR WIRELESS INTERNET

For video streaming application, it is desirable to adjust its transmission rate according to the perceived congestion level in the network to maintain a suitable loss level and fairly share bandwidth with other connections. Furthermore, it’s favorable for the streaming applications to be aware of the transmission quality of wireless channel to obtain good streaming quality by appropriate error protection. With the above considerations and taking the characteristics of wireless Internet environment into account, here we employ the TCP-friendly multimedia streaming protocol for mobile wireless Internet, WMSTFP, which consists of the features as follows.

• Accurate loss differentiation. WMSTFP can accurately detect packet losses caused by errors in wireless channel using the information acquired at the link layer. By jointly using the status information at link layer and the sequence number of incoming packets, we can effectively differentiate different types of packet losses in wireless Internet.

• Forward loss ratio estimation. We observed that packets had different loss patterns, i.e., different loss burstiness length in different types of networks. In this work, we use two Gilbert models to describe the burstiness of these two types of packet losses, respectively. Consequently, we can forwardly estimate packet loss ratio and packet error ratio.

• Smoothed RTT measurement. Since the noisy wireless channel introduces large delay variations, the packet round trip time will fluctuate sharply. Therefore, choosing a single packet as an observation sample can not accurately reflect this variation. In WMSTFP, we propose a method to measure the "average" round trip time during a period of time. As a result, the rate adjustment performs more smoothly while achieving comparable throughput.

In the following sub-sections, we will describe the basic functionalities and processes of WMSTFP in detail.

A. Sender and receiver functionality WMSTFP consists of a sender part and a receiver part. The

sender located in the wired network delivers data packets at a certain rate. The receiver monitors the incoming packets and

feeds back the measured parameters (RTT, loss ratio, etc.) to the sender every round trip time. Based upon the feedbacks from the receiver, the sender adjusts its transmission rate TCP-friendly. The whole process of the protocol consists of the following steps: (1) measuring loss rate (congestive and erroneous); (2) measuring round trip time (RTT) and retransmission time out (RTO); (3) computing available network bandwidth and adjusting the sending rate.

At beginning of the connection, the sender adopts a slow start algorithm to probe available bandwidth. The slow start stops right after a congestive packet loss happens. After slow start, the sender adjusts its sending rate based on the congestive packet loss ratio, RTT and RTO. The estimated packet error ratio is reported to multimedia application for QoS adaptation. During the whole procedure, the receiver measures the incoming packets and estimates the wireless Internet status. Specifically, based on the sequence number of the successfully received packets and the wireless channel status information, the receiver calculates the congestive packet loss ratio and wireless packet error rate. It also gets the timestamp information from every incoming packet and helps the sender to estimate RTT more accurately. The estimated packet loss ratio and the timestamp information are sent back to the sender every round trip time.

B. RTT and RTO estimation In traditional wired-line Internet, the round trip time will

not fluctuate sharply within a relatively small time scale (e.g., during a round). However, this statement cannot hold in wireless Internet due to the link layer ARQ mechanism adopted in wireless links. In most cases, bursty variation and occasional deep fades in wireless channels result in spikes in the observed RTT values.

In order to suppress the spikes in the measured RTTs without causing extra traffic and power consumption in the reverse path, we need to collect as many RTT observations as possible. We store Ts, the time a packet is being sent which is piggybacked in the header of a data packet, and Tr, the time the packet is arriving, in an array at the receiver. The receiver sends a feedback every RTT. By calculating the difference between Tr-Ts of the last packet and Tr-Ts of other packets in this round, the sender can not only measure the RTT value of the last packet triggering the feedback, but also approximate the RTTs of other packets during the round. This method also cancels the clock offset between the sender and receiver so that clock synchronizing between them is unneeded.

Note that the array of (Ts, Tr) is a considerable overhead if it’s piggybacked in the feedback packet. To minimize the traffic in the reverse path, we find out that, to get the final estimated RTT value, all the RTTs in a round are iteratively fed into an EWMA (Exponentially Weighted Moving Average) filter as the traditional TCP protocol does. Mathematically,

nnn RTTRTTRTT )1(1 αα −+= − , (1)

where α is set to 0.75 considering the need of real-time applications, and RTT is the estimated RTT. Expanding (1)

0-7803-8356-7/04/$20.00 (C) 2004 IEEE IEEE INFOCOM 2004

Page 4: Bit Allocation for Scalable Video Streaming over Mobile

with all (Ts, Tr) values, we find the computing result of all the Ts and Tr values can be obtained on the receiver side [14], which can be carried by the feedback packet. Therefore, the minimum overhead in reverse path is achieved. With the estimated RTT, we compute the retransmission time out (RTO) following the instruction in [24]. The sender halves the transmission speed if no feedback has arrived for a period of RTO time.

In addition to the timestamp information, i.e. Ts, the latest estimated RTT value is also contained in the header of data packets to inform the receiver of the time to send feedbacks.

C. End-to-end packet loss differentiation and measurement As stated in Section I, most current end-to-end methods to

differentiate packet losses suffer the flaw of inaccurate loss type differentiation. In this work, we find out that it’s possible to exploit the link layer information in wireless channels to discriminate between the erroneous losses and congestive losses. For example, in the third generation (3G) wireless communication system, the information provided by radio link control layer (RLC) [25] can be used to deduce a packet loss caused by transmission errors. According to the mechanism in RLC layer, IP packets should be fragmented into several RLC frames before transmission and be reassembled on the receiver side. During this period, transmission fail of any RLC frame will result in an assembly failure of the whole IP packet and the consequent drop of this IP packet. Because the assembly failure notification can be obtained on the receiver side, we can use this information to infer whether a packet is dropped due to corruption. In conjunction with the sequence number of the adjacent correctly received packets on the receiver side, we can differentiate erroneous packets from the congestive lost packets.

In addition to the loss type differentiation, it's believed that the loss pattern is also an essential factor that determines the quality observed by users for real-time multimedia applications [22, 26]. One of the most important metrics that characterize loss pattern is burst length of packet loss [26]. To capture the length as well as the frequency of packet losses, we use two Gilbert models to measure the two types of losses [27].

When multiple packet losses occur where both erroneous packet and congestive lost packet co-exist, we can’t know exactly which packets are lost due to congestion and which are lost caused by wireless errors merely based on sequence number and link layer information. Here we use the arrival time of the erroneous packets to derive the distribution of lost packets among the erroneous packets between two back-to-back correctly received packets. The adjacent erroneous packets with larger time interval may contain more congestive lost packets. With this principle, we can identify the distribution of congestive lost packets among the erroneous packets [14].

After the estimation of the congestive packet loss ratio, round trip time (RTT) and retransmit time out (RTO), we compute the available bandwidth as described in [28]. With the available bandwidth, the sender can adapt its sending rate in an additive increase multiplicative decrease (AIMD) manner [14].

IV. LOSS DIFFERENTIATED R-D BASED BIT ALLOCATION FOR SCALABLE VIDEO STREAMING

OVER WIRELESS INTERNET In this section, we introduce a network-adaptive ULP

scheme for scalable video bitstream according to its importance of different layers. An R-D based bit allocation scheme is further proposed to decide how to distribute between source bits and channel protection bits considering the packet loss patterns of both wireless erroneous packet loss and congestive packet loss. A major challenge in bit allocation over wireless Internet is that, in contrast to traditional wired line Internet, the perceived video quality is affected not only by the bandwidth variation and the packet losses due to network congestion, but also the bandwidth fluctuation and the packet losses caused by channel errors.

As mentioned earlier, different packet loss patterns cause different effects on upper applications [22, 26]. By our study, we found out that the packet losses caused by network congestion are independent of those due to wireless errors. Moreover, the packet loss patterns caused by those two reasons are quite different. That is to say, the packet losses due to network congestion and those caused by wireless errors result in different perceived QoS in video streaming application. Thus we need to treat those different loss patterns differently. We use two Gilbert models to measure the patterns of the two types of packet loss. Based upon the two models, we can effectively perform bit allocation taking both the network congestion and the wireless channel error into account. The details will be discussed subsequently.

A. Network-adaptive ULP and loss pattern effects on the perceived video quality PFGS is a multilayer scalable video codec. It stores the

most important information such as motion vector, etc. in the base layer. Other information such as texture is bit-plane coded and stored in several enhancement layers. Different layers of the same frame are correlated in PFGS. To be specific, the higher layer information relies on the corresponding ones in the lower layers. The layer-dependency characteristic of PFGS codec is well suited for prioritized transmission. Since the current Internet only provides best-effort service, prioritized transmission can be achieved by applying ULP scheme to different layers. In our work, ULP is achieved by protecting different layers with different FEC codes. More specifically, strong channel-coding protection is applied to the base layer data stream while weaker channel-coding protection is applied to the enhancement layer parts.

We use Reed-Solomon (RS) codes for FEC channel coding across video packets. An RS code denoted as RS(n, k) represents an n symbol length code which contains k source symbols and n-k protection symbols (redundant data). Generally RS(n, k) can correct t = (n-k)/2 symbol errors. That is to say, if at least n-t out of n symbols are correctly received, the underlying source information can be correctly decoded. Otherwise, none of the lost symbols can be recovered by the receiver.

Now the issue remains is how to allocate target bits between source coding and channel protection coding based on

0-7803-8356-7/04/$20.00 (C) 2004 IEEE IEEE INFOCOM 2004

Page 5: Bit Allocation for Scalable Video Streaming over Mobile

0.20.30.40.5

0.60.7

0.80.9

1 2 3 4 5Packet Loss Burst length (pkt)

Prot

ectio

n ra

tio

Packet Loss Ratio 5%Packet Loss Ratio 3%Packet Loss Ratio 1%

the available network status (e.g., bandwidth, different loss patterns for wireless and wired-line part). In general, when the network is in good status, i.e., packet loss ratio and error rate are low, more bit budget should be assigned for source coding and fewer bits should be assigned for channel coding. On the contrary, when network condition is bad, it's necessary to allocate more bits for channel coding, thus fewer bits should be allocated for source coding. This method enables us to add the necessary level of redundancy in channel coding while providing optimal video quality.

Figure 3. Protection ratios under different burst lengths.

As mentioned above, different loss patterns have different impacts on the perceived QoS quality in video streaming. Next, we experimentally study the effects of the perceptual video quality with different loss patterns in wireless Internet. Let the packet loss ratio of this network be fixed (1%, 3%, and 5% in our study) while the loss pattern, i.e., burst length of packet losses, varies. We add protection data by applying RS code to this video bitstream so that its target residual packet loss ratio decreases to only 0.01%. Figure 3 shows the protection ratio comparison for different loss patterns of a video bitstream over a network. Specifically, when packet loss ratio is 5%, if the burst length of packet losses is 1, 38.9% protection data are needed for video delivery. However, if the burst length of packet losses is 5, 83.0% protection data are needed. This simple study shows that it is quite essential that we consider the different loss pattern effect on perceived video quality when we allocate bits over different types of networks, wireless and Internet, for video over wireless Internet.

Figure 4. Comparisons of the 55th reconstructed frame under different bit allocation schemes.

To further investigate the impact of accurately loss pattern estimation on the perceived video quality for PFGS, we deliver

a video clip, Foreman sequence, with the fixed total target bit rate over the network. The packet loss ratio of this network is 5% and its packet loss burst length is 5. If one does not consider this correlation of packet loss, that is to say, one treats the packet loss randomly distributed, the perceived video quality will be largely degraded compared with the one that takes this loss correlation into account. Figure 4 illustrates the 55th frame of the Foreman sequence delivered by the two schemes. We can easily see that the quality of the right image obtained by the scheme which doesn’t consider the loss pattern is very poor; while a significant better quality can be achieved by using the scheme with the knowledge of packet loss pattern.

B. Loss differentiated R-D based bit allocation Taking different loss patterns in wireless and wired network

into account, in this subsection we present a loss differentiated R-D based bit allocation scheme, which minimizes the expected end-to-end distortion through optimally allocating bits between source data and channel data. The bit allocation problem in here is to find a solution of that how much protection should be added to the source data so that the decoder can efficiently recover the lost data dropped in the intermediate router and the corrupted data dropped by wireless link during transmission while not wasting much bandwidth.

In wireless Internet, the end-to-end distortion DT consists of source distortion Ds and channel distortion Dc. The source distortion is caused by source coding such as quantization and rate control. The channel distortion occurs when a packet loss due to network congestion or errors in a wireless channel happens during transmission. Without losing generality, we can assume that Ds and Dc are uncorrelated, thus the end-to-end distortion can be presented as

),()( cscssT RRDRDD += , (2)

where Ds is a function of source coding rate Rs, and Dc is a function of Rs and channel coding rate Rc. In this work, we address how to optimally distribute bits between source coding and RS coding based upon the above R-D function. Specifically, the solution of bit allocation is periodically performed according to the estimated network bandwidth, RT, obtained by WMSTFP. Then, this problem becomes to allocate the available bit rate such that the optimal Rs and Rc are obtained by minimizing end-to-end distortion under the constraint Rs+Rc≤RT. That is,

. subject to

),()(min},{

Tcs

cscssTRR

RRR

RRDRDDcs

≤+

+= (3)

As PFGS is a layered scalable video codec, it can be truncated anywhere in enhancement layers. Assume the video bitstream can be cut by the source rate Rs and can be stopped at layer l, then the source distortion Ds can be expressed as

)()()(

)(1

11

−−

−−+−

=ll

llslslss RR

DRRDRRRD , (4)

0-7803-8356-7/04/$20.00 (C) 2004 IEEE IEEE INFOCOM 2004

Page 6: Bit Allocation for Scalable Video Streaming over Mobile

where Dl is the distortion when the bitstream is truncated at the end of layer l with rate Rl. Because of the insertion of the resynchronization markers and the similar statistical distortion property of the same enhancement layer [23], the distortion reduction at the same enhancement layer can be linearly calculated approximately.

Note that we consider a wireless Internet connection where only the last hop is wireless. In this case, the channel distortion is caused by two parts: one is caused during the transmission over wired-line part of the connection, Dc,wired, and the other is caused during the transmission over the wireless channel, Dc,wireless. Because the independence of these two distortions, we represent the channel distortion as

),(),(),( ,, cswirelessccswiredccsc RRDRRDRRD += . (5)

PFGS codec adopts bit-plane coding to encode the quantized DCT coefficients. Because the dependency among the bit planes, any corrupted macro-block in the lower layers will cause the corresponding macro-blocks in the higher layers un-decodable. Assume the number of packet needed to be sent in the ith layer is ni, Dc,wired and Dc,wireless can be respectively represented as

,),(),()),(1(),(

and ),(),(),(

1 1,,,

1 1,,

∑ ∑

∑ ∑

= =

= =

××−=

×=

l

i

n

jclayerwlayerccswirelessc

l

i

n

jclayerccswiredc

i

i

jiDjiPjiPRRD

jiDjiPRRD (6)

where Dc(i, j) is the distortion increment caused by the loss of the jth packet in the ith layer. Pc,layer(i, j) is the probability that the jth packet in the ith layer is lost due to congestion while the corresponding packets in the lower layers are correctly received. Pw,layer(i, j) is the probability that the jth packet in the ith layer is lost due to wireless errors while the corresponding packets in the lower layers are all correctly received. Considering the dependency among different layers, Pc,layer(i, j) and Pw,layer(i, j) can be further represented as

,)),(1(),(),(

and )),(1(),(),(

1

1,,,,,

1

1,,,,,

∏ ∏

∏ ∏−

= ∈

= ∈

−•=

−•=

i

u vlayerpacketwlayerpacketwlayerw

i

u vlayerpacketclayerpacketclayerc

u

u

vuPjiPjiP

vuPjiPjiP

ω

ω (7)

respectively, where Pc,packet,layer(i, j) and Pw,packet,layer(i, j) are respectively the probabilities of which the jth packet in the ith layer is lost due to congestion and wireless errors after FEC decoding. We will describe how to calculate these two probabilities later. ωu is the set of the jth packet in the ith layer’s corresponding packets in the uth layer.

By substituting (4) ~ (7) into (3), the R-D based optimal bit allocation for scalable video transmission can be obtained under the given total bit budget, RT. Note that (3) is a non-linear optimization problem with one constraint. We can solve

this problem using optimization methods such as Lagrange multiplier or penalty function method. In this work, we search the global optimized solution by using the policy that takes the following characteristics of PFGS video into account:

• According to the dependency between different layers, we allocate bits to higher layer if and only if all data in the lower layer are allocated.

• Since the lower layer video bitstream has higher importance, the channel protection level of higher layer data can not exceed that of lower layer.

The optimization algorithm is listed as follows.

Step 1: Set the current minimum end-to-end distortion DT_min to a very large value (e.g., 65025 in our work) and allocate all available bit budgets to source data.

Step 2: Find the optimal solution based upon the current distribution of source and channel budget with the constraints in that higher layer protection level can't exceed that of lower layer. Replace DT_min with the current calculated end-to-end distortion DT’ if DT_min > DT’.

Step 3: Decrease the bit budgets assigned to source bitstream. If the current source distortion Ds is larger than distortion DT_min with the solution Rs_min and Rc_min or the source budget decreases to zero, the search ends and the global optimized solutions are found. Otherwise, go to Step 2.

The last problem we are tackling now is how to get Pc,packet,layer(i, j) and Pw,packet,layer(i, j). As mentioned before, to capture the loss patterns of wireless and wired-line networks, we should consider the burstiness of packet loss when calculating the packet loss ratio caused by congestion or wireless errors after FEC recovery. Frossard et al. [29] proposed a method for calculating the packet failure probability after FEC recovery considering the packet loss ratio and average burst length. For video packets protected by RS code RS(n, k), similar to [29], we get

≥+−+•−−

+

+−+•

<

•−++−+•−

•−

+

•++−+•

=

∑∑

∑∑

∑∑∑

∑∑∑

−−−

=

=

+−==

−−

=

−−−

=

−=

+=

+−==

k t )1,1(),()(1

)1,1(),(

k t ),()()1,1(),()(

1),()1,1(),(

),(

1

0

1

1

11

1

1

1

0

1

110

,,

itn

j

k

i

congestion

kn

itj

k

i

congestion

tk

i

itn

j

k

tki

congestionk

ti

kn

itj

t

i

congestion

layerpacketc

knjSkiSikk

p

knjHkiHik

p

kiSikknjSkiSik

kp

kiHiknjHkiHik

p

jiP

(8)

and

≥+−+•−−+

+−+•

<

•−++−+•−

•−+

•++−+•

=

∑∑

∑∑

∑∑∑

∑∑∑

−−−

=

=

+−==

−−

=

−−−

=

−=

+=

+−==

k, t )1,1(),()(1

)1,1(),(

k t ),()()1,1(),()(

1),()1,1(),(

),(

1

0

1

1

11

1

1

1

0

1

110

,,

itn

j

k

i

wireless

kn

itj

k

i

wireless

tk

i

itn

j

k

tki

wirelessk

ti

kn

itj

t

i

wireless

layerpacketw

knjSkiSikk

p

knjHkiHik

p

kiSikknjSkiSik

kpkiHiknjHkiHi

kp

jiP

(9)

0-7803-8356-7/04/$20.00 (C) 2004 IEEE IEEE INFOCOM 2004

Page 7: Bit Allocation for Scalable Video Streaming over Mobile

Bottleneck Link

Wireless link

Base Station

Mobile HostWMSTFP

Senders 1 WMSTFP

Sender n-1 (TCP)

Receiver 2 (TCP)

Receiver n-1(TCP)

Receiver n(TCP)

Sender 2 (TCP)

Sender n (TCP)

……

……

where t = (n-k)/2. Pcongestion and Pwireless are respectively the congestive and wireless packet loss ratio that can be obtained by the two Gilbert models described in Section III.C. H(m,n) and S(m,n) denote the probability that m-1 packet losses occur in the next n-1 packets following one packet loss and the probability that m-1 packets are received in the next n-1 packets following one received packet, respectively, which can be calculated similarly as in [29], therein the Gilbert models are also used.

V. SIMULATION RESULTS In this section, we implement our proposed end-to-end

video streaming architecture and the relevant algorithms. The purpose of this section is to demonstrate that (1) WMSTFP can achieve good performance in wireless Internet. Specifically, it is friendly to TCP in wired line network, and can achieve higher throughput in varying wireless network; (2) our R-D based bit allocation scheme, which differentiates different packet loss patterns in wireless and wired connections, can adaptively adjust source and channel rates to the available estimated bandwidth with good perceived video quality.

Figure 5. Simulation topology.

We use the network simulator (NS) version 2 [30] to study the performance of the WMSTFP protocol and the network adaptive bit allocation scheme for PFGS video streaming over wireless Internet. As shown in Figure 5, we deploy a dumbbell network topology with one congested link to simulate the Internet traffic. The senders reside on one side of the bottleneck link and the receivers are on the other side. All links except for the bottleneck link are sufficiently provisioned to ensure that network congestion only happens at the bottleneck link. All links are drop-tail links. To simulate the dynamically changing wireless environment, we apply a selective ARQ protocol to the wireless link shown in Figure 5. In the wireless channel, all IP packets are first segmented into several RLC frames and then reassembled on the other side. Once the IP reassembly failure on the receiver side occurs, it will be reported to WMSTFP immediately so that it knows an erroneous IP packet was dropped at the RLC layer. In the following simulations, the background traffic is composed of several infinite-duration TCP-like connections.

A. Performance of WMSTFP Here we define TCP-friendly in wireless Internet as the

ability of being friendly to TCP in wired line network and achieving better performance than TCP in wireless part [14]. To get a metrics of the “TCP-friendliness” [24], let kw denote the total number of the WMSTFP connections and kt denote the total number of TCP connections. We also denote the

throughput of WMSTFP connections as wT1 , wT2 , ..., wkw

T and those of TCP connections as tT1 , tT2 , ..., t

ktT , respectively. Then

the average throughputs of WMSTFP and TCP connections are

w

ki

wi

W kTT

w∑ == 1 and t

ki

ti

T kTT

t∑ == 1 , (10)

respectively. The friendliness measure can be defined as

TW TTF /= . (11)

Note that the closer to 1 the value of F is, the friendlier WMSTFP is to TCP.

To study the “rate smoothness” of WMSTFP connections, let 1

iwR , 2iwR , ..., w

i

SwR denote the sending rates at different time

instances 1, 2, ..., Sw of the ith WMSTFP connection and 1ktR ,

2ktR , ..., t

k

StR represent the sending rates at time instances 1, 2, ...,

St of the kth TCP connection, respectively, then the sending rate variation of WMSTFP and TCP connections are, respectively, defined as

∑=

−−=∆w

iii

S

j

jw

jwW RR

1

1 and ∑=

−−=∆t

kkk

S

j

jt

jtT RR

1

1 . (12)

The smoothness measure is then defined as

ki TWS ∆∆= / . (13)

S≤1 means the ith WMSTFP connection is smoother than the kth TCP connection.

Figure 6. Comparisons of throughput for TCP and WMSTFP connections.

To evaluate the TCP friendliness of WMSTFP, we replace the wireless link in Figure 5 with a wired-line link and conduct a simulation under the wired-line networks. In the simulation, we let several WMSTFP connections compete with other TCP connections. We find out that the “friendliness measure” obtained using (11) is 0.98. Figure 6 depicts the simulation results of the throughput for different connections. These

0

500

1000

1500

2000

2500

0 5 10 15 20 25 30 35 40 45 50 55 60

Time (S)

Thro

ughp

ut (k

b) TCPWMSTFP

0-7803-8356-7/04/$20.00 (C) 2004 IEEE IEEE INFOCOM 2004

Page 8: Bit Allocation for Scalable Video Streaming over Mobile

results show that WMSTFP can compete fairly with TCP under congested link.

The smoothness measure calculated using (13) is 0.26 between WMSTFP and TCP. This convincingly shows that WMSTFP can adjust its sending rate in a smoother manner than TCP.

To evaluate the performance of WMSTFP in wireless Internet, we compare the throughput of WMSTFP with that of TCP Friendly Rate Control (TFRC) [12]. Simulation results show that, when RLC frame error rate (FER) are set to 0.1, 0.2 and 0.3, WMSTFP’s throughputs are 6%, 74% and 171% higher than TFRC’s while the packet loss ratios between WMSTFP and TFRC under the same FER only have 1% difference.

We also disable the smoothed RTT estimation method of WMSTFP and compare its rate smoothness, throughput and congestive packet loss ratio with those of the original protocol under different wireless channel conditions. The average rate smoothness measure between the smooth-disabled protocol and the original one is 0.91, and the differences of the throughput and the congestive packet loss ratio between them are 2% and 0.6%, respectively. These results show that, under varying wireless environment, WMSTFP can adapt its transmission rate in a smoother manner while maintaining comparable packet loss ratio and throughput.

B. Performance of loss differentiated R-D based bit allocation scheme To demonstrate the effectiveness of our proposed loss

pattern differentiated R-D based bit allocation scheme over wireless Internet, we conduct the simulation under wireless Internet to test (1) Fixed ULP without loss pattern differentiation over TFRC which is denoted as FULP-T; (2) Fixed ULP without loss pattern differentiation over WMSTFP which is represented as FULP-W; (3) Adaptive ULP over WMSTFP denoted as AULP-W which considers the burstiness of both wireless and congestive packet losses.

We perform simulations under varying bandwidth from 320kbps to 1Mbps and different frame error rate in the wireless channel. The testing video sequence, Foreman, is used in the simulations. Foreman is coded in CIF at a temporal resolution of 10 fps. The first frame is intra-coded and the remaining frames are inter-coded. We use PSNR as a metric to measure video quality. For an eight-bit image with intensity values between 0 and 255, the PSNR is defined as

)/255(log20 10 RMSEPSNR = , (14)

where RMSE stands for root mean squared error. Given an original N×M image f and the processed image f’, the RMSE can be calculated as

[ ]∑∑−

=

=

−×

=1

0

1

0

2),('),(1 N

x

M

yyxfyxf

MNRMSE . (15)

Fair comparison of PSNR among the above three schemes is an important issue. To get a fair protection level for the fixed ULP schemes, we first perform all the simulations of the AULP scheme and get the average protection ratio for base layer and enhancement layers. Then we apply those calculated average protection ratios to the fixed ULP schemes.

Figure 7. Comparison results of average PSNR of different tested schemes under different bit rates for Foreman sequence.

Figure 7 shows comparisons of the average PSNR with different available bandwidth under different RLC FER. It can be seen that our proposed scheme achieves the best performance under different available bandwidth and different quality of wireless channel. From Figure 7 (a), we observe that the higher available network bandwidth, the more efficient the AULP-W scheme comparing with the other two schemes. This is because in our scheme we change the protection level according to the change of current available network bandwidth. Meanwhile, AULP-W considers the loss burstiness for congestive packet losses as well as erroneous packet losses. FULP-W outperforms FULP-T because the TFPC can not differentiate these two kinds of packet losses so that it unnecessarily reduces its sending rate when the packet loss is due to the errors in wireless channel. However, the gain of AULP-W over FULP-W is not so significant when the FER of the wireless channel is 0.2. In this case, the ARQ protocol in the wireless channel can effectively correct most wireless channel errors and the streaming protocol maintains a reasonable sending rate so that the maximum bandwidth utility is obtained while the congestive packet loss ratio is minimized. We can see that in this case the overall packet loss ratio is less than 4%. Under such a low packet loss ratio, we can't expect much gain of AULP-W.

27

28

2930

31

32

33

300 400 500 600 700 800 900 1000

Available Bandwidth (kbps)

PSN

R

AULP-WFULP-WFULP-T

(a) FER = 0.3

2829303132333435

300 400 500 600 700 800 900 1000

Available Bandwidth (kbps)

PSN

R

AULP-WFULP-WFULP-T

(b) FER = 0.2

0-7803-8356-7/04/$20.00 (C) 2004 IEEE IEEE INFOCOM 2004

Page 9: Bit Allocation for Scalable Video Streaming over Mobile

Table I depicts the comparison results of the average PSNR for the whole sequence using those three tested schemes. It can be seen that the gain of the average PSNR of AULP-W over FULP-W is about 1.0dB when FER is 0.3 and 0.4dB when FER is 0.2. The gain of AULP-W over FULP-T is about 2.4dB in both cases.

TABLE I. PSNR COMPARISON RESULTS FOR FOREMAN USING DIFFERENT PROTECTION SCHEMES UNDER DIFFERENT FER

FER=0.3 320kbps 480kbps 640kbps 800kbps 960kbpsAULP-W 28.684163 29.958773 30.874298 31.581641 32.258838FULP-W 28.58115 29.163827 29.552004 30.293835 30.780962FULP-T 28.018016 28.131335 28.538054 27.98044 28.30042 FER=0.2 320kbps 480kbps 640kbps 800kbps 960kbpsAULP-W 29.896785 31.539206 32.537112 33.223339 33.857063FULP-W 29.713924 31.214287 32.160342 32.59875 33.322332FULP-T 28.997481 29.563558 29.88481 30.272214 30.562365

Figure 8. The PSNR comparisons for Foreman using different tested schemes at 720kbps.

Figure 8 shows the comparison results of PSNR for the test sequence Foreman at 720kbps available bandwidth using different schemes. We can see that, along the whole sequence, the AULP-W scheme outperforms the other schemes in most cases. Because we protect the lower layers stronger according to the packet loss patterns as well as the packet loss ratios, the lower layers are less likely to be corrupted during transmission. While for the other schemes, lower layer may be lost when the loss burst length is longer under the same packet loss ratio. It may result in very poor quality and sometimes lead to decoder crash.

Figure 9 shows the reconstructed frames of Foreman sequence using AULP-W, FULP-W and FULP-T. From top to bottom, the left column is the reconstructed 26th frame using AULP-W, FULP-W and FULP-T, respectively, while the right column is the reconstructed 34th frame using the three schemes.

It can be seen from Figure 7 to Figure 9 that our proposed network-adaptive loss pattern differentiated bit allocation scheme obtains better results than the other two schemes in wireless Internet both subjectively and objectively.

Figure 9. Comparison of the reconstructed frames under different FERs. The images in the top, middle and bottom row are reconstructed by AULP-W,

FULP-W, and FULP-T, respectively.

VI. CONCLUSIONS In this paper, we address the issue of how to effectively and

robustly streaming scalable video over mobile wireless Internet. We first present the end-to-end streaming protocol for video streaming over wireless Internet from network QoS adaptation point of view to achieve good performance in both wireless and Internet connections. Then an R-D based bit allocation scheme is proposed from media adaptation point of view to obtain good perceptual video quality. The network adaptation and media adaptation together with a scalable video codec are integrated into a streaming architecture for effective and robust video streaming over wireless Internet. The main contributions of this paper are summarized as follows.

(1) The streaming protocol, WMSTFP, which is friendly to TCP in wired line IP network, and can achieve higher throughput than TCP in wireless network, is an important

1820222426283032343638

0 20 40 60 80 100

Frame Number

PSN

R

FULP-WFULP-TAULP-W

(a) FER = 0.3

24262830323436384042

0 20 40 60 80 100

Frame Number

PSN

R

FULP-WFULP-TAULP-W

(b) FER = 0.2

(a) FER = 0.3 (b) FER = 0.2

0-7803-8356-7/04/$20.00 (C) 2004 IEEE IEEE INFOCOM 2004

Page 10: Bit Allocation for Scalable Video Streaming over Mobile

module in the end-to-end streaming architecture. Since WMSTFP can effectively differentiate congestive packet losses from erroneous packet losses in an end-to-end manner and can suppress the spikes in the round-trip-time values due to the highly varying wireless environment, it can be efficiently adopted in wireless video streaming.

(2) A loss differentiated R-D based bit allocation scheme is further proposed by applying the network-adaptive ULP scheme. It can periodically distribute the available bits between the source bits and the channel protection bits according to dynamically estimated available network bandwidth and different packet loss patterns in wireless and wired line network. The expected end-to-end distortion can be minimized using our proposed loss pattern differentiated bit allocation scheme.

The simulation results demonstrate that our proposed scheme can achieve better results than the one without proper network adaptation and than the scheme which doesn't consider the different packed loss patterns from different types of networks.

ACKNOWLEDGMENT We would like to thank Dr. Feng Wu in Internet Media

group at Microsoft Research Asia for providing the source code of PFGS for the simulations.

REFERENCES [1] M. Civanlar, A. Luthra, S. Wenger, and W. Zhu, “Introduction to the

special issue on streaming video,” IEEE Trans. Circuits Syst. Video Technol., vol. 11, pp. 265-268, Mar. 2001.

[2] C. Chen, R. Lagenduk, A. Reibman, and W. Zhu, “Introduction to the special issue on wireless communication,” IEEE Trans. Circuits Syst. Video Technol., vol. 12, pp. 357-359, June 2002.

[3] Bakre and B. Badrinath, “I-TCP: indirect TCP for mobile hosts,” in Proc. 15th International Conference on Distributed Computing Systems (ICDCS), Vancouver, May 30 - June 2 1995, pp. 136-143.

[4] K. Ratnam and I. Matta, “WTCP: an efficient mechanism for improving TCP performance over wireless links,” in Proc. 3rd IEEE Symposium on Computer and Communications (ISCC '98), Athens, Greece, June 1998, pp. 74-78.

[5] H. Balakrishnan, V. Padmanabhan, S. Seshan, and R. Katz, “A comparison of mechanisms for improving TCP performance over wireless links,” IEEE/ACM Trans. Networking, vol. 6, pp. 756-769, Dec. 1997.

[6] G. Cheung and T. Yoshimura, “Streaming agent: a network proxy for media streaming in 3G wireless networks,” in Proc. IEEE Packet Video Workshop, Pittsburgh, April 2002.

[7] Y. Pei and J. Modestino, “Robust packet video transmission over heterogeneous wired-to-wireless IP networks using ALF together with edge proxies,” in Proc. European Wireless Networks, Florence, Italy, Feb. 2002.

[8] S. Cen, P. Cosman, and G. Voelker, “End-to-end differentiation of congestion and wireless loss,” in Proc. ACM Multimedia Computing and Networking, San Jose, CA, Jan 2002, pp.1-15.

[9] D. Barman and I. Matta, “Effectiveness of loss labeling in improving TCP performance in wired/wireless Network,” in Proc. 10 IEEE Intl. Conf. Network Protocols (ICNP), Paris, France, Nov. 2002, pp. 2-11.

[10] J. Liu, I. Matta, and M. Crovella, “End-to-end inference of loss nature in a hybrid wired/wireless environment,” Boston Univ., Boston, MA, Comput. Sci. Dept., Tech. Rep., Mar. 2002.

[11] S.Biaz and N.Vaidya, “Discriminating congestion losses from wireless losses using inter-arrival times at the receiver,” in Proc. IEEE Symposium on Application-Specific Systems and Software Engr. And Techn., Richardson, TX, Mar. 1999, pp. 10-17.

[12] S. Floyd, M. Handley, J. Padhye, and J. Widmer. Equation based congestion control for unicast applications. [Online]. Available: http://www.aciri.org/tfrc.

[13] J. Widmer, R. Denda, and M. Mauve, “A survey on TCP-Friendly congestion control,” IEEE Mag. Network, vol. 15, pp. 28-37, May/June 2001.

[14] F. Yang, Q. Zhang, W. Zhu and Y.-Q. Zhang, “An End-to-End TCP-Friendly Streaming Protocol for Multimedia over Wireless Internet”, in Proc. IEEE International Conference on Multimedia and Expo (ICME), Baltimore, July 6-9, 2003, pp. II429-II432.

[15] W. Li, “Overview of fine granularity scalability in MPEG-4 video standard,” IEEE Trans. Circuits Syst. Video Technol., vol. 11, pp. 301-317, Mar. 2001.

[16] T. Zhang and Y. Xu, “Unequal packet loss protection for layered video transmission,” IEEE Trans. Broadcasting, vol. 45, pp. 243-252, June 1999.

[17] J. Goshi, R. E. Ladner, A. E. Mohr, E. A. Riskin, and A. Lippman, “Unequal loss protection for H.263 compressed video,” in Proc. Data Compression Conference (DCC), Snowbird, Utah, Mar. 2003, pp. 73-82.

[18] G. Cheung and A. Zakhor, “Bit allocation for joint-source channel coding of scalable video,” IEEE Trans. Image Processing, vol. 9, pp.340-356, Mar. 2000.

[19] U. Horn, B. Girod, and B. Belzer, “Scalable video coding for multimedia applications and robust transmission over wireless channels,” in Proc. 7th Intl. Workshop on Packet Video, Brisbane, Mar. 1996, pp. 43-48.

[20] J. Hagenauer, T. Stockhammer, C. Weiss, and A. Donner, “Progressive source coding combined with regressive channel coding for varying channels,” in Proc. 3rd ITG Conference on Source and Channel Coding, Munich, Jan. 2000, pp. 123-130.

[21] G. Wang, Q. Zhang, W. Zhu, and Y.-Q. Zhang, “Channel-adaptive unequal error protection for scalable video transmission over wireless channel,” in Proc. SPIE Visual Communications and Image Processing (VCIP), San Jose, CA, Jan. 2001, pp. 648-655.

[22] W. Jiang and H. Schulzrinne, “Modeling of packet loss and delay and their effect on real-time multimedia service quality,” in Proc. 10th Intl. Workshop on Network and Operating Systems Support for Digital Audio and Video (NOSSDAV), Chapel Hill, North Carolina, June 2000.

[23] F. Wu, S. Li, and Y-Q. Zhang, “A framework for efficient progressive fine granularity scalable video coding,” IEEE Trans. Circuits Syst. Video Technol., vol. 11, pp. 332-344, Mar. 2001.

[24] Q. Zhang, W. Zhu, and Y.-Q. Zhang, "Resource Allocation for Multimedia Streaming Over the Internet", IEEE Tran. On Multimedia, Vol. 3(3), pp. 339-355, Sept. 2001.

[25] 3GPP Technical Specification 25.322 v5.2.0, “Radio Link Control (RLC) protocol specification,” Sept. 2002.

[26] R. Koodli and R. Ravikanth, “One-way Loss Pattern Sample Metrics,” RFC 3357, Aug. 2002.

[27] H. Sanneck and G. Carle, “A framework model for packet loss metrics based on loss runlengths,” in Proc. SPIE/ACM Multimedia Computing and Networking (MMCN), San Jose, CA, Jan. 2000, pp. 177-187.

[28] J. Padhye, V. Firoiu, D. Towsley and J. Kurose, “Modeling TCP throughput: a simple model and its empirical validation,” ACM/IEEE Trans. Networking, vol. 8, pp. 133-145, Apr. 2000.

[29] P. Frossard, “FEC Performance in Multimedia Streaming,” IEEE Communications Letters, vol. 5, pp. 122-124, Mar. 2001.

[30] Network simulator 2 (Ns-2) [Online]. Available: http://www.isi.edu/nsnam/ns/.

0-7803-8356-7/04/$20.00 (C) 2004 IEEE IEEE INFOCOM 2004