Study On The Key Techniques Of Near-infrared Non-invasive Blood Glucose Detection

With the improvement of people’s living quality, diabetes has become a chronic disease affecting People’s Daily life, and the blood sugar detection method brings necessary physical pain and psychological pressure for patients. Therefore, non-invasive detection method becomes a hot research topic at home and abroad, at the same time, the near infrared spectrum technology and the rapid development of chemometrics method, which widely promoted the near infrared spectrum technology in blood glucose and medical field. This paper mainly discusses the key technical problems of near infrared noninvasive blood glucose detection, the main research content is as follows:1 Studied the theoretical basis of near-infrared spectral analysis, based on the near infrared spectrum of glucose molecules and water features, analyzed the movement of photons using monte carlo simulation methods which finded out the general trajectory of photon in the skin tissue, blood sugar of diffuse reflection detection scheme is proposed.2 Studied the spectrum signal pretreatment method such as Fourier transform, short-time Fourier transform. Considering the respective defects Fourier transform, short-time Fourier transform, the wavelet transform is introduced. In order to improve the computing efficiency of wavelet, lifting wavelet transform theory is analyzed in detail, and put forward the lifting wavelet plan based on the minimum mean square error of adaptive prediction operator. At the same time the commonly used in the field of spectral signal pretreatment method of Savitzky-Golay smoothing method is analyzed. Comparing two methods in Matlab simulation, lifting wavelet adaptive method is superior to the commonly used in the spectrum signal pretreatment method, which more can retain the useful information of the spectrum and has high signal-noise ratio.3 Studied the modeling method of glucose molecules, firstly introduces the method to collect the glucose solution spectrum data and analyzes the selection of the most effective wavelength range when the glucose molecules is modeling. And this paper analyzes the application of partial least square method in model calibration, in allusion to the anti-interference ability of the sample set of problems such as poor and computation speed is relatively slow, puts forward the improving partial least squares method, which is a wavelet partial least squares method, and puts forward the back propagation artificial neural network based on genetic algorithm optimization. The measured spectral data of glucose is established simulation model, using the proposed two glucose correction model and common model SG smooth and partial least squares analysis. The simulation results show that the wavelet partial least squares and back propagation neural network model based on genetic algorithm were higher than SG smooth and partial least squares model commonly used.4 According to human physiological characteristics and the characteristics of the near infrared spectrum of glucose molecules designed a glucose signal extraction module, which is mainly based on the MSP430 chip of TI company, detailed analysis and design of light path and circuit part of the module and each functional module of the software process.