血糖浓度的准确检测对于糖尿病的治疗具有重要的意义。 本文采用连续幂系数回归方法有效地提高了近红外无创检测人体血糖浓度的预测精度。 该方法是传统偏最小二乘法(PLS)的扩展, 实现简易, 且当幂系数选取恰当时, 能够明显地提高预测精度。 应用该方法分别建立了四成分葡萄糖实验和人体口服葡萄糖实验的定量分析模型, 并利用该模型对预测集样本进行预测。 实验结果表明, 与PLS相比, 该方法建立的定量分析模型不仅可以提高预测精度, 而且可以针对不同的测量对象设定不同的幂系数以达到最佳的建模效果。 根据不同个体灵活地选取幂系数, 对于人体血糖浓度近红外无创检测研究具有很大的应用价值。

Accurate measurement of human blood glucose concentration is very significant for the treatment of diabetes. In the present paper, the method of continuum power regression can improve the predictive accuracy of noninvasive measurement of human blood glucose concentration with near infrared spectroscopy. This method is the expansion of the traditional method of partial least squares (PLS). It can achieve simpleness, and can significantly improve predictive accuracy when the power coefficient is fit. Using the method, quantitative analysis models of four ingredient experiment and noninvasive experiment of body were established, and these models can be used to predict the predictive samples. Experimental results show that compared with the PLS, the quantitative analysis models of this method not only can improve predictive accuracy, but also can set different power coefficient for different individuals to achieve the best results of models. According to different individuals, the power coefficient can be selected flexibly, which is of great value to the research on noninvasive measurement of human blood glucose concentration with near infrared spectroscopy.