考虑时间相关性的功耗估计熵算法  

An Improved Entropy to Estimate Power for Considering Temporal Correlation at High-Level

作　　者：吴一品[1] 周锋[1] 陈春鸿 严伟[1] 李舜[1] 彭云峰[1] 

机构地区：[1]复旦大学专用集成电路与系统国家重点实验室,上海201203 [2]Department of Electrical and Computer Engineering,University of Windsor

出　　处：《电子学报》2007年第5期933-936,共4页Acta Electronica Sinica

基　　金：国家自然科学基金(No.90307008)

摘　　要：熵估计是一种在高层次估计功耗的方法,但已经提出的熵算法无法考虑输入信号在时间上的相关性.本文提出了改进熵的概念,在传统熵中加入条件翻转因子,使改进后的熵能够有效估计时间上有相互关联性的信号的翻转率.理论证明和大量BENCHMARK实验结果都表明我们提出的改进熵算法具有合理性和可靠性.The traditional entropy is an efficient method for high-level power estimation,but it doesn＇t work when the input signals are temporal correlated, as is always the case for video and audio streams. This paper aims at this problem. We put forward a new definition of entropy. With the help of the conditional transition probabilities, the proposed algorithm can bring us the estimations with adequate accuracy for temporal correlated inputs. The theoretical proofs and the BENCHMARK experimental results verify the efficiency of our algorithm.

关 键 词：功耗估计 熵 时间相关性 

分 类 号：TP331[自动化与计算机技术—计算机系统结构] TN431[自动化与计算机技术—计算机科学与技术]