PERFORMANCE EVALUATION OF CROSS-LAYER DESIGN WITH DISTRIBUTED AND SEQUENTIAL MAPPING SCHEME FOR VIDEO APPLICATION OVER IEEE 802.11E
AuthorDawoud, Heba Dawoud
MetadataShow full item record
The rapid development of wireless communication imposes several challenges to support QoS for real-time multimedia applications such as video stream applications. Researchers tackled these challenges from different points of view including the semantics of the video to achieve better QoS requirements. The main goal of this research is to design a UDP protocol to realize a distributed sequential mapping scheme (DSM) with a cross-layer design and evaluate its accuracy under different network conditions. In DSM, the perceived quality of a multi-layer video is addressed by mapping each video layer into channel resources represented as queues or access categories (ACs) existing in IEEE 802.11e MAC layer. This research work further investigates the efficiency of this scheme with actual implementation and thorough simulation experiments. The experiments reported the efficiency of this scheme with the presence of different composite traffic models covering most known traffic scenarios using Expected Reconstructed Video Layers (ERVL) and packet loss rate as accuracy measures. This research work also investigates the accuracy of calculating the ERVL compared to its value using actual readings of layers drop rate. The effect of changing the ACs queue size on the ERVL is studied. The use of this scheme shows zero-drop in the base layer in almost all scenarios where no ongoing traffic is presented except that the testing video sessions between nodes. In these experiments, the ERVL continuously reported high values for the number of expected reconstructed video layers. While these values dramatically vary when introducing ongoing different composite traffic models together with the testing video sessions between nodes. Finally, a 40% increase in the ACs queue size shows significant improvement on ERVL while an increase of the queue size beyond this value has very little significance on ERVL.
- Computer Science & Engineering [49 items ]