一种平滑的可变步长算法

本文从长时相关性(LRD)和突发性两个方面,阐明图像组(GOP)时间尺度上的流量能够体现原有视频帧序列的流量特性。并在固定步长自适应算法(FSSA)和可变步长自适应算法(VSSA)的基础上提出了一种新的平滑的可变步长自适应算法(S-VSSA),这种算法可以在GOP所表示的大时间尺度上预测MPEG-4视频流量。仿真实验结果显示:该算法收敛速度快,预测误差低,可以明显改善预测性能。

A Smoothing Variable Step Size Algorithm

In this paper,both Long Range Dependence(LRD)and burstiness are illustrated to show traffic characteristics of orig- inal frame sequence can be represented by traffic on the time scale of GOP.A new smoothing variable step size adaptive(S-VSSA)algo- rithm is presented based on fix step size adaptive(FSSA)algorithm and variable step size adaptive(VSSA)algorithm,which are used to predict MPEG-4 video traffic on large time scale of GOP.And S-VSSA algorithm can improve performance of prediction significantly according to a number of simulations.The new algorithm can achieve fast convergence and small prediction error.