LZ复杂度算法中的二值化方法分析及改进

Analysis and improvement on LZ complexity binary quantifying method

LZ复杂度被越来越多地用于非线性信号分析中 为帮助理解其概念,首先以一个形象的例子概述了LZ复杂度的物理意义,进而作者针对其计算过程中经常使用的粗粒化预处理过程,提出了一种二值化快速实现方法.该方法将原信号减去均值后直接取其二进制补码的符号位作为二值化数值,使计算简单快捷.分析了通常的二值化方法对某些信号产生过分粗粒化的原因,提出了以信号拟合曲线替代均值做为分界的二值化方法,从而有效克服了对基线漂移的信号进行二值化处理时所产生的过分粗粒化问题 用此方法对实际检测的5s时间段SD大鼠脑电信号进行二值化处理,结果表明,该方法不仅计算简单。

LZ(Lem-Ziv) complexity is widely applicable in nonlinear signal analysis. To help understanding its concept, the physical meaning of LZ complexity is elucidated with an example. Concerning with the pre-coarsening process often used in calculating the complexity, a fast binary quantification method is proposed which directly takes the sign bits of the complemental code of the original signal minus its mean as the bi-quantified values, thereby enabling the calculation simple and fast. More important, commonly used binary quantification may cause the so-called over-coarsening. Based on the (analysis) the proposed method is further improved which takes the signal fitting curve instead of its mean as the bi-quantification boundary. Thus the over-coarsening problem caused by signal base line drifting in coarsening process is overcome efficiently. Using this method, 5 seconds SD mouse EEG signal was bi-quantified. The result shows that this method is simple and able to completely eliminate over-coarsening in the commonly used binary quantification.