An NN approach for MPEG video traffic prediction

In this paper, an application of neural networks is suggested for MPEG video traffic prediction. The aim is to predict the future frame of video stream. In single frame prediction problem, we have used the information of previous frame sizes to predict the next frame size of the sequence. As a tool for the prediction, we use multilayer perceptron feedforward neural network. (FMLP).The purpose of our work is to construct suitable neural networks and evaluate the performance of the network for variable bit rate video prediction. We use video traces from. The results of designed predictor indicate that if the networks are trained with a portion of a high activity level trace, it can predict the traffic of other traces of same movie with a reasonable degree of accuracy.