基于液滴指纹图的波形分析算法的改进

摘要

不同的液体在相同的夜滴分析系统中,光纤信号的强弱不同,电容信号的强弱即液滴体积不同,液滴指纹图的波峰波谷的高度、位置以及峰的形状不同,液滴指纹图的曲线长度和曲线下的面积不同;因此,对光电液滴指纹图进行波形分析,计算波形参数,可以实现特征提取并区分不同液体.针对波形分析算法在特征提取过程中存在的两个缺陷,即最值检测时可能错误识别波谷位置和邻域比较时产生大量等值数据集,提出了改进的波形分析算法;对改进前后的波形分析算法进行了分析比较,改进后的算法有效提高了波形参数的准确性.%There are some visible differences for different liquids in the same experimental system, such as the output tiber voltage ( or the light intensity), the output capacitor voltage (or the drop volume), the peak height and the profile of fingerprint, and the curve area surrounded by the fingerprint and the axis. So waveform analysis method and calculated parameters characterizing waveform are employed for feature extraction to identify different liquids. The classical waveform analysis algorithm in feature extraction has two defects: failure to identify the real hollow by extreme detection from maximum and creating excessive equivalent data sets by adjacent comparison. This paper presents an improved one, which improves the accuracy of the waveform parameters.