摘要
根据近红外无创血糖检测技术,设计一种小型近红外无创血糖检测系统。以系统检测探头对人体食指采取透射方式探测,由光电传感器采集四路不同波长近红外光的脉搏波信号。提出一种脉搏波预处理方法,采用经验模态分解和三次样条插值算法去除原始脉搏波的高频噪声和基线漂移。运用动态光谱频域提取法提取对数脉搏波的基波分量。采用偏最小二乘法交叉验证的方法,预测人体血糖浓度。对实验结果进行分析,此系统预测血糖值与真实血糖值的相关性达到了86.5%,预测均方根误差为0.56 mmol/L,证明系统检测结果精度较高。
According to near-infrared non-invasive blood glucose detection technology,miniature near-infrared non-invasive blood glucose detection system is designed. Use system detection probe to detect index finger adopting mode of transmission,collect pulse wave signal of four paths different wavelengths near-infrared light by photoelectric sensor. A pulse wave preprocessing method is proposed,which uses empirical mode decomposition and three times spline interpolation method to delete high frequency noise and baseline drift. Fundamental component of logarithmic pulse wave is extracted by dynamic spectrum frequency domain method. Adopt cross validation method based on partial least square to predict human blood glucose. Analysis on experimental results,correlation of system predicted and real blood glucose reach 86. 5 %,predicted RMSEP is 0. 56 mmol / L,precision of system is verified to be high.
出处
《传感器与微系统》
CSCD
2016年第5期103-106,共4页
Transducer and Microsystem Technologies
基金
国家自然科学基金资助项目(61271119)
国家科技支撑计划资助项目(2013BAI03B01)
广西自动检测技术与仪器重点实验室主任基金资助项目(YQ14116)
广西高等学校科学研究项目(KY2015YB096)
关键词
近红外光谱
光电容积脉搏波
无创血糖检测
脉搏波预处理
偏最小二乘法
near infrared spectroscopy
photoplethysmography
non-invasive blood glucose detection
pulse wave preprocessing
partial least square method
