Professional Documents
Culture Documents
Video Streaming
Video Streaming
: Measurement Based Channel-Adaptive Video Streaming for Mobile Devices over Mobile WiMAX 171
Abstract — The channel bandwidth variation and the 64-QAM offers various data rates to the SS according to the
disconnection during handoff are the most critical problems distance between the base station (BS) and the SS. In addition, the
which degrade the video quality in wireless video streaming. sudden disconnection due to handoff between BSs or sectors leads
To cope with these problems, we propose an efficient video to errors in several frames because the error occurred in one frame
streaming method in this paper, which does not only would be propagated to the subsequent frames due to the prediction
dynamically adjust the video transmission rate based on the of the inter mode, which degrades the video quality significantly
channel bandwidth, but also minimize the error propagation [3],[4].
during handoff. Firstly, the channel bandwidth of the mobile
SS1 BS SS2
worldwide interoperability for microwave access, called
WiMAX, is estimated by analyzing channel parameters Video Input Display
including the carrier to interference and noise ratio (CINR)
Video Encoder Video Decoder
and the handoff is detected by using the handoff occurrence RTP
(Video Bitstream)
message (HOM). Secondly, the streaming server adjusts the Next Transmission
Bitrate
Channel
Parameters
next transmission rate according to the estimated channel Transmission Rate RTCP Channel Bandwidth
bandwidth to avoid the network congestion and performs the Control (Channel Bandwidth) Estimation
intra refresh method that inserts an intra frame (I-frame) right Intra Refresh RTCP Hanoff Detection
Insert (HOM)
after handoff by using the HOM to reduce the error I-frame
Streaming Server Streaming Client
propagation effectively. Experimental results indicate that the
proposed method can improve the performance of the video Fig. 1. Concept of the proposed channel-adaptive video transmission
streaming over mobile WiMAX1.
To address these problems, several methods for wireless
Index Terms — Wireless video streaming, channel-adaptive, video streaming have been proposed [5], [6]. The method
handoff, mobile WiMAX. proposed in [5] adjusted the transmission rate to the varying
throughput of wireless 3G network. However, it requires the
I. INTRODUCTION bandwidth estimation of wireless network and needs to consider
Due to the explosive growth of the wireless multimedia the channel bandwidth in the mobile WiMAX to adapt the video
communication services, there are increasing demands on real-time transmission rate dynamically. In [6], the periodical random
video streaming over the wireless systems. Recent advances in intra refresh and motion information-based conditional intra
high-speed networks have made it feasible to provide real-time refresh methods were proposed to reduce the error propagation
video streaming. Among the advanced wireless standards, in error-prone channel. However, it is not an efficient way to
WiMAX is an emerging wireless communication system that reduce the error propagation caused by handoff latency in
provides high-data rate as well as long-range coverage [1], [2]. The wireless network.
higher quality and seamless streaming in video transmission over In this paper, we propose a channel-adaptive video streaming
the wireless network require to cope with the problems such as method over mobile WiMAX, as shown in Fig. 1, which does
channel bandwidth variation, handoff, transmission error. Among not only dynamically adjust the video transmission rate based
those problems, the channel bandwidth variation and the handoff on the channel bandwidth, but also minimize the error
due to movement of the subscriber station (SS) are the most critical propagation during handoff. Firstly, the current channel
problems. The channel bandwidth variation causes the network bandwidth is estimated by using channel parameters including
congestion when the video transmission rate exceeds the channel the CINR. The estimated channel bandwidth is then exploited to
bandwidth. In case of the mobile WiMAX, the adaptive determine the next video transmission rate, thereby avoiding the
modulation and coding (AMC) scheme from half-rate QPSK to 5/6 network congestion. Secondly, an efficient intra refresh method
is proposed that inserts I-frame right after handoff by using the
1
This research was supported by SAMSUNG ELECTRONICS CO., LTD. HOM to reduce the error propagation effectively.
established by Network Adaptive Video Transmission Project. The rest of the paper is organized as follows. Section II
Hye-Soo Kim, Hyeong-Min Nam, Jae-Yun Jeong, and Sung-Jea Ko are presents the proposed measurement based channel-adaptive
with the School of Electrical Engineering, Korea University, Seoul, Korea
(email: hyesoo@ieee.org, min@dali.korea.ac.kr, jyjeong@dali.korea.ac.kr, video streaming algorithm. Experimental results and
sjko@korea.ac.kr). conclusions are given in Section III and Section IV, respectively.
Soo-Hyung Kim is with the Telecommunication R&D Center, Samsung
Electronics, Suwon, Korea (email: soohyung.kim@samsung.com).
Authorized licensed use limited to: VELLORE INSTITUTE OF TECHNOLOGY. Downloaded on July 22, 2009 at 07:32 from IEEE Xplore. Restrictions apply.
172 IEEE Transactions on Consumer Electronics, Vol. 54, No. 1, FEBRUARY 2008
CAIR RTCP
) APP
) ) ) ( R DL )
(
R =min RUL , R DL )
CARC Insert I-frame
)
R,Target Bit Rate RTP
(Video Bitstream)
Video Encoder Video Decoder
Channel bandwidth For the WiMAX using the orthogonal frequency division
CINR
5/6 64 QAM multiple access (OFDM) and time division duplex (TDD), the
physical data rate, RPHY (Mbps), is obtained using the system
2/3 16QAM
parameters [7] in Table I as follows:
1/2 QPSK
R PHY = D rate ⋅ bmod ⋅ c rate ,
Handoff (1)
⎧1.8432 (Msubcarriers/sec), if uplink,
D rate = ⎨
MS location
⎩ 3.6864 (Msubcarriers/sec), if downlink,
Fig. 3. Relationship of MCS scheme, physical data rate, and CINR
according to the distance between SS and BSs.
where Drate is the data subcarrier rate when the ratio of the
II. PROPOSED CHANNEL-ADAPTIVE VIDEO STREAMING downlink (DL) to the uplink (UL) is 2, bmode is the modulation
gain that bmode = 2 for QPSK, bmode = 4 for 16-QAM, and bmode
Fig. 2 shows the proposed algorithm which consists of the
= 6 for 64-QAM, and crate is the coding rate (bits/subcarrier).
channel adaptive rate control method and the channel adaptive
For example, if the current modulation and coding scheme
intra refresh method.
(MCS) is 5/6 64-QAM, then the physical data rate (RPHY_DL)
A. Channel Adaptive Rate Control (CARC) for DL is equal to 18.432 (=3.6864·(5/6)·log264) Mbps. Fig. 3
In video streaming over wireless network, the available shows that the CINR is effectively used to estimate bmode·crate
bandwidth (AB) on the end-to-end path including both wired in the proposed algorithm.
and wireless links is determined by the wireless link capacity Next, we introduce a novel method of estimating the AB. In
since the wireless link commonly provides lower capacity than the mobile WiMAX network, RPHY is the theoretical upper
the wired one in most cases. We assume that the AB is affected bound of the AB because the AB is influenced by various
only by the capacity of mobile WiMAX networks. For network parameters such as the channel resource overhead,
TABLE I the number of users, and the weighting factor related to the
)
SYSTEM PARAMETERS different environments of the SS. Therefore, the AB, R , for
System Parameters Value both the DL and the UL is estimated as follows:
Frequency band 2.3 GHz
Channel bandwidth ) (1 − Poverhead ) ⋅ (1 − C overhead ) ⋅ W env
9 MHz
R= ⋅ R PHY , (2)
Duplex TDD / 5 msec N user
DL and UP ratio 2:1
Multiple access OFDMA where Poverhead represents the ratio of the packet overhead,
Cell coverage 1 Km Nuser denotes the number of users which is equal to 2 in our
experiments, and Wenv is the weighting factor related to the
Authorized licensed use limited to: VELLORE INSTITUTE OF TECHNOLOGY. Downloaded on July 22, 2009 at 07:32 from IEEE Xplore. Restrictions apply.
H.-S. Kim et al.: Measurement Based Channel-Adaptive Video Streaming for Mobile Devices over Mobile WiMAX 173
environment of the SS. Coverhead in (2) is the ratio of the NGOP_org NGOP_org NGOP_org
channel resource used for overhead to total channel resource
in one frame considering the mobile WiMAX frame structure P P P P P I P P P P P I P
[7]-[9] as follows: …… ……
Authorized licensed use limited to: VELLORE INSTITUTE OF TECHNOLOGY. Downloaded on July 22, 2009 at 07:32 from IEEE Xplore. Restrictions apply.
174 IEEE Transactions on Consumer Electronics, Vol. 54, No. 1, FEBRUARY 2008
Home TABLE II
AAA
PHYSICAL DATA RATE WITH CINR LEVEL
Home
ACR Agent Coding RPHY_UL
WiBro Core CINR (dB) Modulation RPHY_DL (Mbps)
Network Rate (Mpbs)
Router
WiBro Coverage
RAS1 26 64-QAM 5/6 18.43 -
Internet
RAS2 23 64-QAM 3/4 16.59 -
20 64-QAM 2/3 14.74 -
PSS1 (Streaming Server)
ACR
18 16-QAM 5/6 12.29 6.14
ACR: Access Control Router
PSS2 (Streaming Client) RAS: Radio Access Station
PSS: Portable Subscriber Station
16 16-QAM 3/4 11.05 5.53
AAA: Authentication , authorization and accounting,
* BS: Base Station (ACR + RAS) 14 16-QAM 2/3 9.83 4.92
* MS: Mobile Station (PSS)
TABLE III
WiBro NETWORK PARAMETERS
PCMCIA Card
Parameters Value
(a) (b)
ThRSSI -60 dB
Fig. 6. Snapshot of the experimental test platform. (a) Environment of
Thhandoff 0 dB
field test. (b) Implemented video streaming system.
Poverhead 0.026 (= 40 Bytes / 1500 Bytes)
)
where B GOP is the estimated number of bits for the new GOP, F 0.7 for LOS
) Wenv
is the frame rate, R is the estimated AB using (4), and NGOP_new 0.4 for NLOS
is the size of incoming GOP after handoff. The bit allocation
scheme using (7), the CAIR algorithm can achieve the best video
quality under a given channel condition.
6000
DL AB per user
III. EXPERIMENTAL RESULTS 5500 UL AB per user
5000
In this section, the performance of the proposed algorithm is 4500
4000
of the IEEE 802.16 standard and is developed based on the 3000 2 users 1 user
mobile WiMAX [15]. Fig. 5 illustrates the experimental 2500 3 users
2000 4 users
environment in the WiBro network which consists of the
1500
portable subscriber station (PSS), the radio access station (RAS), 1000
the access control router (ACR), and so on. In Fig. 5, the PSS 0 20 40 60 80 100 120 140 160
and the RAS are the same as the SS and the BS of the mobile Time (sec)
WiMAX network, respectively. Fig. 6 shows the snapshot of the
experimental platform. In Fig. 6 (a), the PSS is connected to the Fig. 7. Maximum AB per user according to the number of users. (a)
WiBro network in the 60Km/h speed vehicle. Fig. 6 (b) shows Maximum DL AB per user. (b) Maximum UL AB per user.
Authorized licensed use limited to: VELLORE INSTITUTE OF TECHNOLOGY. Downloaded on July 22, 2009 at 07:32 from IEEE Xplore. Restrictions apply.
H.-S. Kim et al.: Measurement Based Channel-Adaptive Video Streaming for Mobile Devices over Mobile WiMAX 175
-40 -40
-50 -50
-55
-55
RSSI (dB)
RSSI (dB)
-60
-60
ThRSSI = -60 dB ThRSSI = -60 dB
-65
-65
-70
-70
-75
0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100
Time (sec) Time (sec)
(a) (b)
Fig. 8. Measured RSSI of each link. (a) Measured DL RSSI. (b) Measured UL RSSI.
6000 30
12000 30
8000 20 4000 20
Bandwidth (Kbps)
Bandwidth (Kbps)
CINR (dB)
CINR (dB)
6000 15 3000 15
4000 10 2000 10
2000 5 1000 5
0 0 0 0
0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100
Time (sec) Time (sec)
(a) (b)
Fig. 9. Estimated ABs for each link. (a) Estimated AB for DL. (b) Estimated AB for DL.
Fig. 9 shows the comparison between the estimated AB “FOREMAN” sequence with 1880 frames of CIF format is
using (2) and the actual AB. As shown in Fig. 9, the proposed encoded into an MPEG-4 bitstream with the video
AB estimation method well tracks the variation of the actual transmission rates of each method. As shown in Fig. 10 (c),
AB of both the DL and the UL. Therefore, the streaming the CARC method avoids the network congestion while the
server with the CARC algorithm correctly determines the next constant bitrate (CBR) method generates the network
video transmission rate according to (4). congestion around 20, 60, and 85 sec. It is seen that even
Fig. 10 (a) shows the variation of CINR and the though the average transmission rate of the proposed CARC
corresponding HOMs when Thhandoff is equal to 0. In Fig. 10 method (= 1.49 Mbps) is greater than the one of the CBR
(a), HOMs generated at 21 and 64 sec indicate the DL handoff method (= 1 Mbps), the proposed CARC method avoid the
and the UP handoff, respectively. It is seen that the proposed network congestion effectively as shown in Fig. 10 (c). To
approach using (6) well detects the handoff of both DL and evaluate the CAIR algorithm, the PSNR comparison is
UL. Thus, the CAIR algorithm can insert the I-frame right performed between the proposed CAIR algorithm and the
after the handoff. method proposed in [6], namely random intra refresh (RIR).
In the RIR method, I-frame is inserted when the motion
activity of a frame is greater than a certain threshold. In our
For the performance evaluation of the proposed methods simulation, the number of the inserted I-frames in the RIR
including the CARC and the CAIR methods, we perform the method is 30. In Fig. 10 (c), the CAIR effectively reduces the
PSNR comparison of each method. In Fig. 10 (b), the video error propagation caused by handoff around 21 and 64 sec,
transmission rates for the each method are shown as compared while the RIR does not.
with the estimated AB for both DL and UL. In Fig. 10 (c), the
Authorized licensed use limited to: VELLORE INSTITUTE OF TECHNOLOGY. Downloaded on July 22, 2009 at 07:32 from IEEE Xplore. Restrictions apply.
176 IEEE Transactions on Consumer Electronics, Vol. 54, No. 1, FEBRUARY 2008
35
Measured DL CINR
Measured UL CINR
30
25
20
CINR (dB)
15
10
HOM HOM
5
Thhandoff = 0 dB
0
10 20 30 40 50 60 70 80 90
Time (sec)
(a)
7000
Estimated AB for DL
6000 Estimated AB for UL
Video transmission rate for CARC
5000 Constant bit rate (CBR)
Transmission rate (Kbps)
4000
3000
2000
1000
10 20 30 40 50 60 70 80 90
Time (sec)
(b)
55
CBR
50
CBR+RIR
CAIR RIR
45 CARC
CARC+CAIR
40 CAIR
35
PSNR (dB)
30
25
20
Fig. 10. Channel conditions and PSNR comparison. (a) Variation of the CINR and HOM. (b) Target bit rate. (c) PSNR comparison.
Authorized licensed use limited to: VELLORE INSTITUTE OF TECHNOLOGY. Downloaded on July 22, 2009 at 07:32 from IEEE Xplore. Restrictions apply.
H.-S. Kim et al.: Measurement Based Channel-Adaptive Video Streaming for Mobile Devices over Mobile WiMAX 177
TABLE IV
PERFORMANCE COMPARISON
Test Sequence Method Average PSNR [dB] σ of PSNR [dB] Packet loss ratio (PLR)
CBR 26.63 10.40 0.219
CBR+RIR 28.04 9.97 0.212
FOREMAN
CARC 35.08 5.89 0.022
CARC+CAIR 36.60 3.93 0.023
CBR 30.05 10.13 0.220
CBR+RIR 34.08 9.95 0.214
NEWS
CARC 36.72 6.49 0.025
CARC+CAIR 38.99 5.41 0.027
CBR 19.80 6.53 0.219
CBR+RIR 21.48 6.30 0.222
MOBILE
CARC 24.56 4.98 0.025
CARC+CAIR 27.24 3.18 0.024
of the mobile WiMAX network, but also minimize the error [14] Q. Chen, Z. Chen, X. Gu, and C. Wang, “Attention-based adaptive intra
refresh for error-prone video transmission,” IEEE Trans. Commun.
propagation due to the handoff latency. The combination of Mag., vol. 45, pp. 52–60, Jan. 2007.
the CARC and the CAIR enables the proposed video [15] S. Q. Lee, N. Park, C. Cho, H. Lee, and S. Ryu, “The wireless broadband
streaming to minimize the video quality degradation in the (wibro) system for broadband wireless internet services,” IEEE Trans.
Commun. Mag., vol. 44, no. 7, pp. 106–112, Jul. 2006.
mobile WiMAX network. Experimental results show that the
proposed method provides the seamless and high quality video
streaming over the mobile WiMAX network.
Hye-Soo Kim received the B.S. and M.S. degrees in
REFERENCES electronics engineering from the Department of
[1] K. H. Teo, Z. Tao, and J. Zhang, “The mobile broadband WiMAX Electronics Engineering at Korea University, in 2002 and
standard,” IEEE Signal Processing Mag., vol. 24, pp. 144– 148, Sept. in 2004. He is now the Ph.D candidate in the Department
2007. of Electronics Engineering at Korea University. His
[2] IEEE 802.16 TGe Working Document, (Draft Standard) - Amendment research interests are in the areas of Fast Handoff, IEEE
for physical and medium access control layers for combined fixed and 802.11, IEEE 802.16, 3G cellular network, NGN, Mobile
mobile operation in licensed bands, 802.16e/D4, Aug. 2004 QoS, IP QoS, video signal processing and multimedia
[3] F. Ohrtman, WiMAX handbook, Building 802.16 wireless networks. New communications.
York, NY: McGraw-Hill, Jun. 2004.
[4] G. Dong and J. Dai, “An improved handover algorithm for scheduling
services in IEEE8O2.16e,” IEEE Proc. Mobile WiMAX Symposium, pp.
25–29, Mar. 2007. Hyeong-Min Nam received the B.S. degree in
[5] J. W. Kim, H. M. Nam, S. J. Lee, J. Y. Lee, and S. J. Ko, “Probing-based Electronics Engineering from Korea University in 2004.
channel adaptive video streaming for wireless 3G network,” IEICE He entered the Multimedia Communication Processing
Trans. Communications, vol. E89-B, no. 2, pp. 357-363, Feb. 2006. Lab in the Department of Electronic Engineering of
[6] P. Haskell and D. Messerschmitt, “Resynchronization of Motion Korea University, in 2005. He is now a Ph.D. candidate
Compensated Video Affected by ATM Cell Loss,” ICASSP ’92, San at Korea University. His interests are wireless video, QoS,
Francisco, CA, vol.3, pp. 545–48. Mar. 1992. video signal processing, and multimedia communication.
[7] S. Cherry, “WiMax and Wi-Fi: separate and unequal,” IEEE Spectrum,
vol. 41, no. 3, pp. 16–16, Mar. 2004.
[8] S. Shin, C. K. Kang, J. C. Kim, and S. H. Oh, “The Performance
Comparison between WiBro and HSDPA,” IEEE Proc. Wireless
Communication Systems, pp. 346–350, Sept. 2005. Jae-Yun Jeong received the B.S. degree from Korea
[9] R. Lee, J. Kim, J. Yu, J. Lee, and D. Kim, “Capacity analysis University in the Department of Electric Electronics and
considering channel resource overhead for mobile Internet access Radio, in 2005. He entered the Multimedia
(WiBro),” Proc. the IEEE Int. Conf. Vehicular Technology Conference, Communication Processing Lab in the Department of
pp. 3082–3086, Jun. 2005. Electronic Engineering of Korea University, in 2005. He
[10] WiMAX’s technology for LOS and NLOS environments, [online is now a Ph.D. candidate at Korea University. He has
available]:http://www.wimaxforum.org/technology/downloads/WiMAX currently focused on Wireless QoS, video signal
NLOSgeneral-versionaug04.pdf processing and multimedia communication processing.
[11] Y. Kim, J. Y. Pyun, H. S. Kim, S. H. Park, and S. J. Ko, “Efficient real-
time frame layer rate control technique for low bit rate video over
WLAN,” IEEE Trans. Consumer Electron., vol. 49, pp. 621– 628, Aug.
2003. Soo-Hyung Kim received the diploma in electrical
[12] S. F. Yang and J. S. Wu, “Handoff management schemes across hybrid engineering from the Kwangwoon University, Korea in
WiMAX and Wl-Fi networks,” Proc. the IEEE Int. Conf. TENCON, pp. 1998. Since joining Samsung Electronics in 1998, he has
1–4, Nov. 2007. worked as engineer in various multimedia transmission
[13] J. W. Kim, S. Y. Jeon, H. S Kim, J. W. Yun, and S. J. Ko, “Seamless related research, standardization. He currently is a
Video Streaming Technique Supporting Error Concealment for VOD research engineer whose main area of interest is
Services in Vertical handoff,” IEICE Trans. Fundamentals., vol. E89- multimedia service delivery over mobile networks. The
A, no. 6, pp. 1571-1577, Jun. 2006. focus of his current work is on mobile broadcast services.
Authorized licensed use limited to: VELLORE INSTITUTE OF TECHNOLOGY. Downloaded on July 22, 2009 at 07:32 from IEEE Xplore. Restrictions apply.
178 IEEE Transactions on Consumer Electronics, Vol. 54, No. 1, FEBRUARY 2008
Authorized licensed use limited to: VELLORE INSTITUTE OF TECHNOLOGY. Downloaded on July 22, 2009 at 07:32 from IEEE Xplore. Restrictions apply.