الفهرس 
Previous work
VOIP TECHNOLOGYOnly 14 pages are availabe for public view
 |  | from | 150 |  |  |
| | | from | 150 | | |
Abstract
Voice over Internet Protocol (VoIP) is one of the fastest growing technologies in communication networks in the world. New multimedia Applications have a critical requirement on quality of service (QoS). The quality of service is mainly affected by network deficiencies such as jitter, packet loss, and delay. Network jitter as a variable delay is a dramatic problem in communication networks especially in Voice over IP network (VoIP). An advanced contribution is to have an adaptive scheme to cope with the variable network conditions.
In this thesis the optimal use of artificial intelligence as an effective tool for taking smart decisions is proposed as a solution to adapt the playout time. First an algorithm was coded using MATLAB program to analyze network state in different conditions, i.e. propagation delay, variance, initial delay in the network. Several network behavior patterns were extracted from these simulations to represent a diverse jitter responses. Next, we introduce an adaptive approach using Back Propagation (BP) neural network model to identify the jitter and adjust the playout time according to extracted network patterns. The Neural network model was designed to have the network conditions as an input and predicts the suitable playout algorithm to be applied according to these conditions.
The algorithm was tested using k-fold cross validation and repeated k-fold cross validation methods. The algorithm shows a great efficiency in adjusting the playout time to avoid packet drop. After that, another novel approach has been proposed in this thesis using fuzzy logic to manage the jitter problem effectively. The fuzzy logic model was designed to give more flexibility in decision making concerning the playout time. An algorithm have been developed and tested against different network conditions patterns. The results show that the algorithm can achieve promising results under different delay conditions. Finally, a comparison has been made to evaluate the performance of the two approaches and a state of all results was demonstrated.