Page 211 - ITU Kaleidoscope 2016

ADAPTIVE VIDEO STREAMING OVER HTTP USING
STOCHASTIC BITRATE PREDICTION IN 4G WIRELESS NETWORKS

Dhananjay Kumar, S. Aishwarya, and A. Srinivasan, and L. Arun Raj

Department of Information Technology, Anna University, MIT Campus, Chennai
dhananjay@annauniv.edu, aishwarya.gsv@gmail.com, asriniit@gmail.com, arun4u85mit@gmail.com

ABSTRACT match the available bit rate in the network. The traditional
streaming method based on progressive download fails to
Video streaming over Hypertext Transfer Protocol (HTTP) cope up with dynamic network traffic [2] thereby degrading
used in multifarious applications creating a multimedia the media quality.
environment faces a challenge in 4G wireless network due to The streaming techniques [3] are classified into three major
the fluctuating nature of internet traffic and variable classes: i. traditional (Real-time Transport Protocol (RTP) /
capacity of wireless channel over time. The existing Dynamic RTP Control Protocol (RTCP)), ii. progressive streaming
Adaptive Streaming over HTTP (DASH), though works well (HTTP/TCP), and iii. adaptive streaming (HTTP/TCP,
for stored video up to some extent, poses a complication in UDP). The HTTP based Adaptive Streaming (HAS) has
live transmission thereby depreciating the streaming quality exhibited resilience to the internet traffic and hence widely
due to high link bit rate fluctuation. In this paper, we have used as DASH [4] in the present systems. The use of DASH
proposed an efficient ARIMA Based Bit Rate Adaptation in entertainment based utilities, where the stored videos are
(ABBA) model in the receiver/client side that estimate the being streamed to the client, requires segment based
link traffic based on the incoming packet bit rate to predict information and pre-defined streaming parameters to
the subsequent future link capacity in order to notify the facilitate ease in deciding the upcoming bandwidth changes.
sender/server. Based on the response from the receiver the However in live streaming where the video content is created
server adapt its outgoing stream as per forecasted link data and encoded only when the systems connect in real time over
rate, and hence eliminate the degradation of video due to the network, the adaptability of DASH to intimate the sender
channel throughput variations. The proposed ABBA about the link bandwidth becomes an encumbrance for
algorithm was implemented on IP over 4G wireless network targeting an improvement in the perceived quality of video
and the streaming quality was evaluated on several full by the user [5].
reference metrics of video quality. The test result For a case study on existing 4G wireless network, the uplink
outperformed an existing buffer based approach and also a and downlink data rate on the Airtel 4G LTE-TD Hotspot [6]
fuzzy based adaptation algorithm. For example, the ABBA system was monitored in laboratory environment (Fig. 1).
algorithm exhibited an average increase of 22 % in PSNR Although these wireless systems are designed to support up
and 9% in SSIM than a buffer based method. to 100 Mbps in high mobility access, the actual capacity at
user premises not only fall much below the specified values,
Keywords: 4G Wireless, HTTP, ARIMA, Client-server, but also fluctuate over time. Clearly there is a high incentive
Adaptation, Video quality evaluation metrics in developing a video streaming system which can adapt to
this network operating environment.
1. INTRODUCTION

The user data traffic in wireless mobile network has been
increasing rapidly across the globe. As per the Cisco Visual
Networking Index [1], the 4G wireless network will have the
highest stake (40.5 %) of total mobile connections
worldwide, and 75% of the global mobile data traffic will be
video by 2020. Such remarkable progress is fueled by the
video streaming service over internet by YouTube, Netflix,
etc. The ever increasing number of smart phones with
internet access over 4G wireless network is another reason
for the tremendous increase in streaming video traffic.
The ultimate objective of all streaming services is to deliver Figure 1. Bitrate observed during streaming of live videos
seamless content to the end user in real-time, though it poses using Airtel 4G dongle [6]
a huge challenge due to large fluctuating bandwidth in the
network. To provide user the seamless multimedia service The variations of incoming bit rates while the video is being
with maximum achievable Quality of Experience (QoE), the streamed can be analogized to a time series to rigorously
media content in particular video need to be adaptive to analyze the past observations to make forecasts about

978-92-61-20431-0/CFP1668P-ART © 2016 ITU – 193 – Kaleidoscope