摘要
提出了一种新的基于时间序列分析的改进型MPEG视频流量模型,利用消除趋势项、滑动平均过程等方式建模,并根据实际流量的分布,将预测流量进行概率分布转换.大量的仿真实验等方面说明,在预测误差、尾部概率分布、自相关函数和自相似性及GOP、视频帧两个时间尺度上,该模型同时兼顾了长时相关性和短时相关性,并保持了与实际流量相一致的自相似性.
A novel improved MPEG video traffic model based on time series analysis presented by using trend component elimination and MA process in modelling. The frame_size probability distribution of predicted traffic cwas transformed under real situation. The prediction error, tail of probability distribution, auto_correlation function and self_similarity were considered. The model takes account of both LRD and SRD on GOP and viedo frame, and agrees with real traffic in self_similarity.
出处
《北京交通大学学报》
CAS
CSCD
北大核心
2004年第5期32-35,共4页
JOURNAL OF BEIJING JIAOTONG UNIVERSITY
基金
国家自然科学基金资助项目(60472034)