Fuzzy Scene-Change Based Real-Time Traffic Prediction for H.264 VBR Video Sources
Prediction of real-time bursty VBR video traffic is critical to dynamic network resource allocation and management. Many previous studies have focused on prediction of VBR video traffic involving complex techniques like Neural Networks, Wavelets, Multi-Fractals, etc., but real-time prediction techniques are useful only when they are simpler, faster and accurate. Moreover, during scene changes, prediction accuracy is drastically reduced. Thus, the authors present here a hybrid, fuzzy based, real-time traffic predictor for H.264 VBR traces that adapts to scene change fluctuations. Their focus is on predicting I-frame traffic, as they are the key frames of any standard MPEG/H.264 based encoders.