A Error Correction Algorithm for Video Streaming over Internet
Bo Tian, Yimin Yang, Shuting Cai
School of Automation
Guangdong University of Technology
Guangzhou, China
e-mail: tianbomail@163.com
Abstract—The compressed video streaming delivery over a
best effort network such as Internet is sensitive to packet loss
and time delay, which can severely damage end-to-end video
quality. To address this problem, a predicted packet loss rate
based error correction algorithm for video streaming is
proposed in this paper. We rely on a simple hidden Markov
model based mechanism to predict packet loss rate. The probe
gap model is exploited to estimate the available bandwidth.
Based on the estimated packet loss rate and available
bandwidth, quantization parameter is adjusted to meet the
bandwidth, and the forward error correction or automatic
repeat request was adaptively selected to recover from the
channel errors. Simulation results demonstrate that the
proposed algorithm reduce the probability of packet loss rate
and achieve a higher average video peak signal-to-noise ratio
compare to the hybrid automatic repeat request algorithm.
Keywords-video transmission; Internet; packet loss rate;
error correction
I. INTRODUCTION
With the video applications grow rapidly, transmission of
real-time video streaming over the best-effort Internet is one
of the most challenging issues. A major problem is that the
video streaming are very sensitive to end-to-end
performance problem like packet loss rate and time delay,
which can severely damage the quality of the received video.
Therefore, the design of reliable and robust video
transmission over Internet is a significant challenge [1].
Forward error correction (FEC), automatic repeat request
(ARQ) and hybrid FEC and ARQ (HARQ) are three
commonly used error correction algorithms. FEC usually
utilize Reed-Solomon (RS) code to encode the source
packets. However, the FEC lack the capability of adapts to
time-varying channel and lead to serous waste of bandwidth
in the case of heavy packet loss rate[2]. The ARQ has less
redundancy packets as compare to FEC, but it requires a
feedback channel and its mechanism of error control is more
complicated. In addition, ARQ induced more transmission
delay due to its retransmission mechanism [3]. In [4],
analysis and experiment results showed that the HARQ can
provide better error resilience performance, as long as
reasonable adjust the FEC parameters based on predicting
the packet loss rate accurately. Address to this problem, Li
and Fernando predict packet loss rate statistics with two-
state Markov model, and quantitatively analyzed the impact
of packet loss rate on the video distortion [5], [6], but their
proposed algorithm incurs high implementation complexity
due to high computational complexity.
By deeply studying both algorithms of FEC and ARQ,
we proposed a predicted packet loss based error correction
algorithm (PPLECA) to eliminate the influence of the loss
rate of packets and delay during video transmission over
Internet. At first, we applied the hidden Markov model for
predicting future packed loss rate statistics. Second,
according to the statistics, the FEC or selective ARQ was
adaptively selected to perform error correction on the video
streaming. The optimal redundancy of RS code has been
calculated and the maximum number of retransmission for
selective ARQ has been analyzed. Simulations show that the
algorithm efficiently decreases the packet loss rate of video
streaming and provide better video service at receiver.
The remainder of the paper is organized as follows. In
Section II, the hidden Markov model based packet loss rate
prediction algorithm is described. In Section III, the
predicted packet loss rate based error correction algorithm is
proposed. Simulation showing the performance of algorithm
is given in Section IV and finally conclusion is drawn in
Section V.
II.
H
IDDEN MARKOV MODEL BASED PACKET LOSS RATE
PREDICTION
A. Available Bandwidth Estimation
Available bandwidth estimation is a very important
problem to video streaming transmission. When the
estimated available bandwidth is less than a threshold value,
in order to avoid further network congestion and reduce the
probability of packet loss and play off interruption at
receiver, the transmitter increase the quantization parameter
adaptively and decrease the video transmission rate. We
employed the probe gap model proposed in literature
[7] .The available bandwidth is computed as follow:
111
(
estimation
MKN
ii
iii
LM K N
BW
)
i
gg
¦¦¦
(1)
Where is the packet length,
L
,MK
and is the number
N
of packets with increased gaps
i
, unchanged gaps
i
and decreased gaps
i
g
,respectively.
____________________________________
978-1-4673-6278-8/13/$31.00 ©2013 IEEE