一种高可靠性的光电容积脉搏特征提取方法

An optimisation method for the extraction of eigenvalues of optoelectronic volumetric pulse wave

摘要:

光电容积脉搏波可以反映血糖浓度,因此准确提取光电容积脉搏信号对血糖监测具有重要意义。本文提出了一种针对光电容积脉搏特征值提取的优化方法,可以消除局部异常信号进而提高无创血糖检测的精准度。该方法将采集到的脉搏波信号经过经验小波变换(Empirical Wavelet Transform, EWT)去除高低频噪声后,分解成多个子信号,并利用基于动态时间规整算法(Dynamic Time Warping, DTW)对分解后的子信号进行两两相似度判别计算,识别并剔除异常子信号,随后分别建立保留与剔除异常子信号的两种数据集合,最后提取各自的相关特征参数进行对比分析。实验结果表明,相对于保留异常子信号的集合,利用DTW算法剔除异常子信号的集合中所提取的特征参数的峰峰值、上升支、下降支速率以及信号长度数据集合的标准差分别降低了31.6%、14.8%、44.2%以及28.5%,表明利用该方法可以提升特征值提取的可靠性。此外通过利用该优化的特征方法可以将光电容积脉搏波特征值的采集稳定性提高44.4%,证明了该方法在无创血糖监测中的实用性与可靠性。

Abstract:

s: The photovolumetric pulse wave can reflect the concentration of blood glucose, so the accurate extraction of the photovolumetric pulse signal is of great significance for blood glucose monitoring. In this manuscript, we propose an optimized method for extracting the eigenvalues of the photovolumetric pulse, which can eliminate local abnormal signals and thus improve the accuracy of non-invasive glucose detection. The method decomposes the collected pulse wave signal into multiple sub-signals after removing the high and low frequency noises by Empirical Wavelet Transform (EWT). Then the two-by-two similarity discrimination calculation is performed on the decomposed sub-signals via Dynamic Time Warping (DTW) algorithm in order to identify and eliminate the abnormal sub-signals. The two data sets of retained and rejected abnormal sub-signals are established, and the relevant parameters are extracted for comparative analysis. The experimental results show that the standard deviation of the peak-to-peak value, the rate of rising branch, the rate of falling branch, and signal length are reduced by 31.6%, 14.8%, 44.2%, and 28.5%, respectively, via using the DTW algorithm. This indicates that the reliability of the extraction of the feature values can be improved by using this method. In addition, the collection stability of eigenvalues of photovolumetric pulse wave can be improved by 44.4%, which proves the practic ality and reliability of this method in the monitoring of non-invasive blood glucose.