Page 213 - ITU Kaleidoscope 2016

ICTs for a Sustainable World

2.1.2 Receiver Sub-modules time series model that can be used to predict the future trend
considering the current network conditions.
The following modules illustrate how the client analyses
(Fig. 3) the link instability based on ABBA Algorithm. MAIN
a) Playback: The client requests the sender to start streaming
and initiates connection with the sender in an appropriate
port using HTTP.
b) Analyze: The client analyzes the incoming bit rate of PLAYBACK SEND
packets to ascertain how the link will support stream in near Request Send feedback in
future using ABBA algorithm. server for the linked port
c) Send Feedback: The client creates a message based on video
defined format of feedback and makes an intelligent decision Populate the
Establish
to alert the sender. connection using message based
sockets on strategic
decision
MAIN
Create a new Create status
instance of VLCJ code for
INIT ADAPT to play transmission

Capture video Set instance of
ML and stream

Set the media Modify the parameters
player locator Synchronization Compute media Inspect the
based on decision of multi threads statistics using trend and
Init VLCJ Player ABBA forecast the
and set trans Interpret the message Inbound succeeding
coding options after continuous measurement of changes
listening packet bitrate Compute AIC

ANALYSE
Instantiate the LibVlc objects Stream in the linked Figure 3. Client side modules
port with the
media resource are created with requesting client
Player Factory
locator
3.1.1 Auto-regressive (AR) Component

STREAM The Auto-regressive (AR) part is used to establish the
Figure 2. Server side modules covariance between the bit rates fluctuating over time [13]
that can be used to foresee how the variations would take
3. SYSTEM MODEL place in the future.
p
The client side of the proposed system has higher complexity Auto_Reg = −1  φ i L (1)
i
than the server, and it applies stream analysis algorithm to i = 1
handle the non-linearity in the incoming traffic as they may where ϕ i represent covariance and L i lag operator for i th
vary rapidly over time that is too complicated to fit into any packet, and p denotes the number of bit rate samples taken
specific predefined classes. A heuristic based stochastic over time.
algorithm is formulated to overcome the existing problems
and ensure delivery of higher quality videos in the prevailing The covariance ϕ i signifies a statistical relationship between
circumstances. Based on the predicted link behavior the bit rate and time that is used for trend analysis and foresees
client identifies the trend of the series that is labeled as i. the upcoming bit rates, which is expressed as:
advancing, ii. degrading, iii. oscillating, and iv. stable. The φ = E [ X , X s ] μ ×− t μ s (2)
t
traffic load in the network could momentary
increase/decrease or prolong to increase/decrease. The where µ t and µ s represents the mean associated with the
projected incoming flow rate is then sent to the server as a random variables X t and X s.
feedback for it to adapt effectively and modify its parameters
instantaneously. 3.1.2 Auto Regressive Integrated Moving Average (ARIMA)
Model
3.1 Stochastic Prediction Model
Considering X tp to be predicted bit rate where ‘tp’ denotes
There is need to explore a suitable statistical model which the index number, ARMA(p,q) model with the integration of
captures the dynamics of incoming bit rate and maps to a correlation factors is defined as:

– 195 –