Date

2016

Major

Electrical Engineering (B.S.)

Document Type and Release Option

Thesis (open access)

Faculty Mentor

Dr. Rami Haddad

Abstract

In this work, we propose the use of non-linear, autoregressive neural network models for predicting video frame sizes. This model utilizes H.265 encoded video traces as inputs and the predicted future frame sizes as outputs. This model is developed to predict ultra-high definition video frame encoded with H.265 within IP networks. The video I, P, and B frames are predicted separately to improve model prediction accuracy. This approach is verified in MATLAB using various H.265 video traces. The results indicate that the proposed models were able to predict the video traffic fairly accurately.

Share

COinS