Journal of Software Engineering and Applications, 20 12, 5, 8-13
doi:10.4236 /js ea.2012.512b002 Published Online December 2012 (http://www.SciRP.org/journal/jsea)
A Control Mechanism of Stream Media Based on 3G
Network
Wei Jiang, Limin Meng, Songxiang Ying, Hong Peng, Zhijiang Xu
Zhejiang Provincial Key Laboratory of Communication Networks and Applications, College of information Engineering, Zhejiang
University of Technology, Zhejiang, China.
Received 2012
ABSTRACT
In this paper, we use RTP/RTCP over unreliable UDP to realize the transportation of real time streaming. Meanwhile,
according to the transmission feedback and network parameters, we analyze and calculate the network delay, packet
loss and RTT (ro und trip time) to dete rmine the net work state . Finall y, we pr op ose a stre aming media tra ns mission co n-
trol scheme which could sense the network state and quickly adjust the rate of the sending side. It takes congestion,
packet loss rate into comprehensive consideration and improves the overall performance of 3G network stream media.
Keywords: RTP; D e la y; RTT; Co ntrol Mechanis m; 3G
1. Introduction
With the development of wireless network, a variety of
applications based on that come into live. A lot of re-
search has already done on how to improve the perfor-
mance of the high speed wireless network, including
control mechanism. An Explicit Congestion Notification
(ECN) [1] based on congestion control mechanism con-
trols sending rate by marking the IP packets but can not
control rate properly when a continuous packets loss
happens. The end-system based source algorithm detects
the nature and discriminate packet loss types to improve
network performance by relative one-way trip time
(ROTT) [2] or packet inter-arrival times [3].
As the development of the wired network is already
quite mature, most of the current wireless applications
are built in the IP packet network based on heterogene-
ous wireless-wired hybrid network in which two main
error types may affect the quality of stream media when
we transfer media data [4,5]. They are congestion error in
wired network and burst interference error in wireless
channel. Most of the algorithms mentioned above can not
deal with the burst interference error. Several approaches
have been made to differentiate these two types of error
[6-8].
In this paper, we distinguish the wired congestion as
long time congestion from the wireless congestion as
short time congestion. Our focus is to design a rate con-
trol mechanism considering the packet loss rate and RTT
whic h can a dj ust t he se nde r b ehavi or s moo thl y and whe n
congestion happens we can detect, classify and control it,
thus we may get a hig her uti liza tion of the ne twor ks, i m-
prove the throughput and increase the packet successful
delivery rate of the wireless networks.
2. Protocol and System Module
2.1. RTP/RTCP
RTP is an IP based application layer protocol which sup-
ports real time audio and video data transmission. It en-
capsulates the stream application data into RTP packet
and the n trans fers throu gh UDP. It meets t he needs of high
efficiency and real time character in real time stream
media transmission since UDP is an unreliable and con-
nectionless oriented protocol which does not contain re-
transmission mechanism nor waiting for confirmation
mechanism.
RTCP is real time control protocol which, in other
word s, sho uld be used to gether with RT P. Whe n the pro-
gram starts an RTP session, it will occupy two ports for
both RTP and RTCP. RTP itself does not guarantee the
reliabilit y of the data, nor d oes it analyze and co ntrol the
network throughput and congestion. All of this should be
done with RTCP. Generally speaking, RTCP uses the
same distribution mechanism with RTP. It sends control
message to all the session members periodically and re-
ports to application about the related information con-
cerning the session member. The information usually
includes the quantity sent, quantity received, packet loss
rate, network delay, jitter and congestion. The applica-
tion might make full use of th ese kinds o f infor matio n to

A Control Mechanism of Stream Media Based on 3G Network
9
adjust network state for the sake of media QOS (quality
of service). The detailed information of RTP/RTCP is
shown in RFC 3550[9].
2.2. System M odul e
Stream media transmission system is the core of the
whole s ys tem and it s ma i n j o b is t o gua rante e a high effi-
ciency stream media transmission. To achieve that, there
must be an interactive scheme between server and client
which could provide a better QOS under all the circums-
tances. Figure 1 shows the stream media transmission
system module. Front end audio and video data are en-
capsulated into RTP packet which would be stored and
transmitted in the send buffer. This process is regulated
through RTCP in order to prevent the network conges-
tion. While the client receives the RTP packets, it puts
them i nto rece ive buffer for resto ring and analyzi ng. The
restored data is sent to decoder for audio and video and
the analyzed data is built into an RTCP packet to the
sender for feedback.
2.3. Network Par ameters
Black screen, mosaic, image pause, buffering phenomena
often occur in the process of stream media services
which strongly reduce the service QOS. Many factors
contribute to these cases among which network conges-
tion and packet loss are the main causes.
In order to improve the QOS of stream media, we take
network parameters as real time transmission condition
and d ynamica lly cha nging b asis . During t he trans missio n,
stream server gathers RTCP data periodically and calcu-
lates the QOS parameters. Sender receives RR (Receiver
Report) and analyzes the current state of the network,
then promptly adjusts the rate of media by changing the
resolution and frame rate.
The packet loss rate is defined as the ratio of the num-
ber of the missing data packet and sending data packet.
Defin e the packet loss rate and the number of packets
expected . Before the moment t, the receiver got the
maximum and minimum sequence of RTP packet as
and , so here we have:
(1)
Suppose in reality the number we get the RTP packet
is and the lost , then:
(2)
From the above, we get the packet loss rate:
(3)
RTT value calculation:
After the SR arriving, receiver sends an RR for feed-
back. We define the delay between the SR and RR as
DLSR (delay since last SR). We get LSR (last SR time-
sta mp) fro m the NT P of SR a nd p ut LSR a nd DLS R into
the corresponding RR field. Suppose at moment A the
receiver receives SR, then the one-way delay is (A-LSR).
The receiver sends the RR at time B , the n DLS R is B -A.
Finally the sender receives the RR at time TB, then the
RTT value can be calculated:
(4)
3. Stream Media Control Mechanism
The streaming tra nsmission co ntrol is mainly reflected in
the media sending rate adjustment according to the net-
work conditions. We mainly focus on reducing packet
loss rate and the probability of congestion to achieve a
higher quality of video.
Figure 1. Str eam media transmission system mo du le.

A Control Mechanism of Stream Media Based on 3G Network
10
3.1. Image Quality Determination
For long time congestion, packet loss is mainly due to
network overload, so packet loss rate can be used as the
basis of long time congestion. For short time congestion,
large value of RTT is always expected, so we detect RTT
value as the basis of short tim e congesti on.
First, let’s take long time conge stion into conside ratio n.
We define the network parameters’ threshold according
to the image quality requirement. These values include
packet loss rate and jitter etc according to which the
network status can be classified into three categories.
They are under-loading, full-loading and over-loading.
Let M and N be the two threshold values for the three
states. When packet loss rate is less than M, it indicates
that the net work is under-loading, so we can increase the
sending rate to acquire a higher QOS. When packet loss
rate is between M and N, it means the network is
full-loading and we don’t have to change the sending rate.
But when packet loss rate is more than N, we should
slow down the sending rate to convert the over-loading
network to a normal level. Fi g ur e 2 shows the relation-
ship between network status and packet loss rate.
Packet loss rate directly affects the quality of picture
which can be determined by PSNR (Peak Signal Noise
Ratio) and MOS (Mean Opinion Score) [10]. PSNR is
defined as follow:
( )
( )( )
col row
peak
2
00
col row
PSNR
20lg 1,, ,,
NN
SD
ij
n
V
Yni jYni j
NN = =



=

− 



∑∑
(5)
The meani ng of each part is shown in Table 1:
MOS is widely used in audio and video quality mea-
sure ment. T his metho d uses a rithmetic average to get t he
system running status of quantitative indicators of sub-
jective evaluation. MOS can be divided into 5 grades as
grade 5 means the picture is of best quality while grate 1
indicates the picture worst. Table 2 shows the corres-
pondence between PSNR and MOS.
3.2. Determine the Threshold
Experiment uses multiple H.264 video streams which
have been packaged into RTP packets and under multiple
random packet loss condition to determine M and N.
According to formula (3) and (5) we draw a relationship
between packet loss rate and P SNR. Also, we take down
the RTT value of each stream. The experiment is rec-
orded in Table 3:
According to the table, we get Figure 3. F ro m Figure
3 we can see, when PSNR is less than 20 dB, which
means packet loss rate is greater than 3%, MOS value is
2, that is to say, the picture quality is very poor. Gener-
ally speaking, when PSNR is greater than 20 dB, the
video quality is acceptable. On the basis of Figure 3, we
select N 4% and M 2%.
Under-loading Over-loading
Full-loadingPacket loss rate
MN100%0%
Figure 2. Network status and p acket los s rate.
Table 1. Meani ngs of sy mbo l in PSNR.
Signal Meaning
n Picture sequence number
peak
V
peak 21
k
V= −
，k is the number of pixel per bit
( )
,,Yni j
Luminance value of a pixel
col
N
The number of the horizontal pixels
row
N
The number of the vertical pixels
Table 2. Relationship between PSNR and MOS.
PSNR MOS
>37 5(Excellent)
31-37 4(Good)
25-31 3(Fair)
20-25 2(Poor)
<20 1(Bad)
Table 3. Experiment reco rd.
Number of
video RTT value (ms) Packet loss rate PSNR (dB)
1 15 0.8 29.8
2 30 1.3 26.2
3 43 1.7 25.0
4 62 2.6 23.6
5 74 3.3 19.2
6 88 4.2 16.8
7 105 5.0 12.8
8 152 7.5 10.0
9 183 10.0 6.6

A Control Mechanism of Stream Media Based on 3G Network
11
0
5
10
15
20
25
30
35
0246810 12
Packet loss rate/%
PSNR/dB
Figure 3. Relationship between PSNR and packe t loss rate .
Moreover, for the sake of network stability, a packet
loss smoother is used to smooth the packet loss parame-
ter. This is particularly effective when network faces the
jitter problem as we shall observe a rather longer period
of time to determine the packet loss and send rate. After
the smooth procedure, the packet loss rate is:
( )()()( )
11pnpn pn λλ =− −+
(6)
where
01 λ <<
, the larger
λ
is, the more influence
the new packet loss has. Here we set
λ
= 0.3.
4. Rate Control Mechanism
As we have already seen in Table 3, PSNR does not only
have relatio nship with pa cket loss rate but a lso RTT val-
ue. It is shown that if the current RTT value exceeds a
nor mal too muc h, the qualit y shar ply go es do wn, and it ’s
also not proper to have the same control algorithm with
those whose RTT value is at a relative normal value. So
when short time congestion occurs, we reduce the send
rate to half of the current one. When long time conges-
tion happens, we adjust the send rate by AIMD[41]
scheme as the following formula.
( )
()( )
()( )
()( )
1 02%
1 2%4%
1* 4%
Rn pn
Rn Rnpn
Rn pn
α
β
−+< ≤


=−<≤

−>

(7)
where
α
and
β
are constants for rate adjustment, R
(n) is the send rate at n moment, and 2% and 4% is the
threshold value. The whole control mechanism flow
char t is shown in Figure 4.
4.1. Experiment and Result
After we implement the control mechanism, the experi-
ment r e sult i s s ho wn i n Figure 5. Video server can adjust
to network congestion by regulating the send rate through
RTCP. The overall QOS is improved. Figure 5(a) is a
picture whose RTT value exceeds a normal value too
much, which means the network encounters a short time
congestion. So we cut the send rate value to half of the
former. Figure 5(b) shows the result. Figure 5(c) is a
picture whose packet loss rate is 3.2% and PSNR value is
19 dB. It has mosaic effect and MOS grate reaches only
2. Fig ure 5(d) is a picture whose rate has been regulated
and it has a packet loss rate of 1.5%, and its MOS reach-
es 27. There is no mosaic on the picture and the total
quality is acceptable.
Record the RTT
value
start
Is there a new
RR arrive?
Is the current
RTT within a normal
value?
Packet loss
Rate < 2 ?
'RR α = +
' 1/2*RR=
Packet loss
Rate > 4 ?
'*RR β =
Finish?
end
Yes
Yes
Yes
Yes
No
No
No
No
Yes
No
Figure 4. Control mechanism flow chart.
5. Conclusion and Further Work
In this paper, we use RTP/RTCP over unreliable UDP
to realize the transportation of real time streaming over
3G network. Several parameters are measured during the
transfer as packet loss rate and RTT value to determine
the network status. Finally, we propose a stream media
transmission control scheme considering the above two
parameters to regulate the send rate. Experiments show
that this kind of scheme could enhance the network sta-
bility by adj usti ng to network var iatio n.

Current Distortion Evaluation in Traction 4Q Constant Switching Frequency Converters
12
(a) (b)
(c) (d)
Figure 5. Experiment result.
There are several future works in this paper. Since we
have already implemented the algorithm considering
packet loss rate and RTT, some other parameters such as
jitter and bandwith might be under consideration in the
future.
6. Acknowledgement s
This work was supported in part by Zhejiang Provincial
Key Laboratory of Communication Networks and Ap-
plications, College of information Engineering, Zhejiang
Unive r sit y o f T e chno l og y, Program for Zhejiang Leading
Team of Science and Technology Innovation and Appli-
cati ons a nd Zhejiang University of Technology.
REFERENCES
[1] H.J. Lee, H.J. Byun, J.T. Lim, "TCP-friendly congestion
control for streaming real-ti me ap plications o ver wireless
networks," IET Communications, 2008, vol. 2, no. I, pp.
159-163.
[2] N. Nguyen, E.H. Yang, "End-to-end loss discrimination
for improved throughput performance in heterogeneous
networks," In: 2006 3rd IEEE Consumer Commu-
nications and Networking Conference (CCNC
2006), Las Vegas, NV, January 2006, pp. 538-542.
[3] Y. Tobe, Y. Tamura, A. Molano, S. Ghosh, H. Tokuda,
"Achieving moderate fairness for UDP flows by
path-status classification," In: 25th Annual IEEE Confe-
rence on Computer Network (LCN 2000), Tampa, FL,
November 2000, pp. 252-261
[4] Mengyue. Chen, “QoS algorithmic research based on
wir e d -wireless hybrid network transmission,” master
Thesis, Nanjing University of Posts and Telecommunica-
tions, Nanji ng , 2007.
[5] Yang. Xia, “WVTPx transmission rate control algorithm
based on wired-wireless h ybr id network” master Thesis,
Nanjing University of Posts and Telecommunications,
Nanjing, 2007.
[6] S. C en , P .Cos ma n, and G.Voel ker, “End to end differen-
tiation of congestion and wireless lo ss,”i nP roc.ACM mu l-
timedia Computing Networking, SanJo s e, CA,
Jan.2002,pp.1–15.

A Control Mechanism of Stream Media Based on 3G Network
13
[7] D.B arman and I.Matta, “Effectiveness of loss labeling in
improving TCP performance in wired/wireless net-
work,”inProc.10th IEEE Intl. Conf. Net work Protocols
(ICNP),Paris,France,Nov.2002,pp.2–11
[8] J.Liu, I.Matta, and M.Crovella, “End-to-end inference of
loss nature in a hybrid wired/wireless environment,” Pro-
ceedin gs of WiOpt’03: Modeling and Optimization in
Mobile, AdHoc and Wireless Networks, 2003, pp.78–82.
[9] Henning Schulzrinne, Stephen L. Casner, Ron Frederick
and Van Jacobson, “RTP: A Transport Protocol for
Real-Time Applications,” The Internet Society (IEFT),
RFC3550, July 2003.
[10] KE C H, SHIEH C K, HWANG W S, et al. An evaluation
framework for more realistic si mulations of MPEG video
transmission [J]. Journal of Information Science and En-
gineering, 2008,24(2), pp.425-440.