动态光谱信号质量的评估与筛选

摘要

Near infrared spectroscopy analysis,as a kind of nondestructive real-time continuous detection method,has provided ideas for the noninvasive measurement of blood components.In this article,in order to ensure the accuracy and reliability of the collected spectral data,the 405 acquired samples are evaluated by combining valid single edge counts of dynamic spectrum pulse wave in the time domain with the quality factor Q value of dynamic spectrum pulse wave in the frequency domain.As a result, the abnormal samples are removed and 2 1 8 cases of valid samples are selected.We use the dynamic spectrum data of the selected 2 1 8 samples as the experimental group and another 2 1 8 samples as a control group modeling analysis with the hemoglobin con-centration of the corresponding samples.Each group select 200 cases of samples as a calibration set and 18 cases of sample as a prediction set.The prediction accuracy of the experimental group reach 93.8%.The prediction accuracy of the two control group respectively evaluated by the valid single edge counts or Q value are 65.6% and 67.7% and the three unfiltered control groups are 53.7%,33.3%,42.6% respectively.The prediction mean relative error (MSEP)of the experimental group is 0.067 5,the other two control groups are 0.072 3 and 0.072 2,and the other three control groups are 0.082.3,0.078 9,0.082 8.So com-pared with another control groups,the MSEP of the experimental group is the minimal.The results show that the filtering method of the spectral data samples through combining time domain with frequency domain is reliable and effective.this will pro-vide a method to the precision research of dynamic spectrum noninvasive detection.%近红外光谱分析作为一种无损、实时、连续检测的方法,为血液成分无创检测提供了思路。为了保证采集到的光谱数据的准确性和可靠性,本文通过动态光谱脉搏波时域有效单沿数和频域品质因数 Q值相结合的方法对405例采集的样本进行了评估,剔除异常的样本,筛选出218例可靠样本。对筛选出的218例样本和对照组的218例样本的动态光谱数据与相应的被测对象的血红蛋白浓度值进行了建模分析,每组选取200例样本作为校正集,18例样本作为预测集。有效单沿数和 Q值联合筛选的实验组的预测准确率达到93.8%,有效单沿数或Q值单一评估的两个对照组的预测准确率分别为65.6%和67.7%,未经过筛选的三个对照组的预测准确率分别为53.7%,33.3%,42.6%；实验组的预测平均相对误差为0.0675,单一评估的两个对照组分别为0.0723和0.0722,随机选取的对照组分别为0.0823,0.0789,0.0828,较其他各组相比为最小。结果表明时域频域联合法对光谱数据样本筛选具有可靠性和有效性,这将为动态光谱的无创检测的精准性研究提供了方法。