The Pretreatment Methods Of Data In The Study Of Non-Invasive Measurement Of Human Blood Glucose Concentration With Near-Infrared Spectroscopy

Near-infrared spectroscopy analysis technique is efficient, rapid, low cost, noninvasive, not destroying environment. It is not only suitable for laboratory analysis, but also for in-field fast and real-time on-line analysis. In the Non-invasive measurement of human blood glucose with spectroscopy, the Near-infrared spectroscopy analysis technique is the major method. However, with the complex components of human blood, the precision of the instrument, and the changes of the measure environment, there is much noise in the spectra. That will affect the prediction precision and robustness of the model. A way of improving the robustness is the pretreatment methods. With the pretreatment methods, the spectra-to-noise ratio was greatly improved while the noise was suppressed effectively. They also reduce the spectrum difference of the same samples under different instrument and different conditions.This article introduces the cases of domestic and external theory of Fourier analysis and Wavelet analysis. Firstly, it analyzes and evaluates the two basic theories in detail. Secondly, it also studies the multivariate regression method and the abstraction of interesting spectral signal and overcome of multi-collinear with it. At last, we applied the PLS (Partial Least Square) regression method after the Fourier transform and wavelet transform to the Near-infrared spectroscopy analysis. The result shows that, it can reduce model complexity with preserved prediction ability, and abstract predictive information for predictor variable from spectrum effectively and can achieve a robust model.