中文摘要：

近红外无创血糖浓度测量面临着信号微弱、仪器信噪比低、血液中其他成分干扰以及人体生理环境不断变化等因素影响,致使血糖信号难以辨识这一难题。浮动基准法针对葡萄糖浓度变化对吸收和散射效应的影响,选择吸收效应和散射效应相互抵消的位置获得基准光谱,在吸收和散射效应引起最大综合影响的位置获得测量光谱,并采用基准光谱对测量光谱进行修正,从而消减样品背景状态变化以及仪器噪声、漂移等干扰。通过应用实验评估浮动基准法对提高模型预测精度和稳定性的有效性。通过对处理前后的预测结果进行比较分析得出：应用浮动基准法数据处理之后,交互验证均方根误差（RMSEP）最大改进比率34.7%,实验结果表明浮动基准法能够有效消减样品自身状态变化以及仪器噪声、漂移等干扰因素的影响,较大幅度地提高了模型的预测性能和稳定性,以期突破无创血糖浓度测量的障碍。

英文摘要：

Weak signal,low instrument signal-to-noise ratio,continuous variation of human physiological environment and the interferences from other components in blood make it difficult to extract the blood glucose information from near infrared spectrum in noninvasive blood glucose measurement.The floating-reference method,which analyses the effect of glucose concentration variation on absorption coefficient and scattering coefficient,gets spectrum at the reference point and the measurement point where the light intensity variations from absorption and scattering are counteractive and biggest respectively.By using the spectrum from reference point as reference,floating-reference method can reduce the interferences from variation of physiological environment and experiment circumstance.In the present paper,the effectiveness of floating-reference method working on improving prediction precision and stability was assessed through application experiments.The comparison was made between models whose data were processed with and without floating-reference method.The results showed that the root mean square error of prediction（RMSEP） decreased by 34.7% maximally.The floating-reference method could reduce the influences of changes of samples＇ state,instrument noises and drift,and improve the models＇ prediction precision and stability effectively.