Study On The Extraction Method Of Dynamic Spectrum

The non-invasive blood components measurement as the research hotspot in the biomedicalfield has been deserving special attention. In this paper, the existing non-invasive detectiontechnologies are firstly presented and analyzed, and the method of near infrared spectroscopy(NIRS) analysis is considered to have obvious technological superiority, which can be divided intotraditional NIRS analysis and issue background interference elimination analysis. By the formermethod, the emission spectra of the monitoring site are directly used for data modeling andprediction; while by the latter method the spectral information of blood is extracted from the totalemission spectra before the modeling analysis; so the latter method can effectively eliminate theinfluence of issue background and shows a great potential of development.The method of dynamic spectrum (DS), proposed by our group, belongs to one of themethods eliminating the interference from the issue background. DS method can theoreticallyobtain a pure absorption spectrum of pulsating blood by utilizing the Photoplethysmography(PPG), but the spectral signal of pulsating blood is weak and easily influenced by measuringcondition, and the limited accuracy of an instrument is another influential factor. One of theeffective measures for resolving the above problems is making full use of the advantage of digitalsignal processing to optimize the extraction method of DS. So this paper focuses on the extractionmethods of DS, and the main works and innovations are as follow:(1)For improving the low precision of traditional peak-peak extraction method, the FourierTransform (FT) extraction method and single-trial extraction method were put forward earlier, andthen the research was focused on the different value (D-value) extraction method and theextraction method based on a fast digital lock-in amplifier, and some exploratory researches onroot-mean-square and envelope-area extraction method were presented. In this paper, the basicprinciple, the implementation steps and the features of each method were introducedcomprehensively from two directions---frequency domain and time domain.(2)In order to deeply analyze the performance of each method, a simulation experiment wascarried out under the specific application of measuring oxygen saturation. A data acquisitionsystem of dual-wavelength pulse waves was built to obtain a standard dual-wavelength pulsewaves, then the simulation signal disturbed by different interferences were constructed based onthe standard signal. By processing these simulation signals, the performances of each method wereeffectively evaluated. (3)On the basic of simulation study, the practical application of these extraction methods wasresearched. The initial data of DS were processed by each method to get the correspondingnormalized DSs. By comparing the discrete distribution level and smoothness degree of thesenormalize DSs, the actual performance of each method was evaluated objectively, which couldprovide guidance for the subsequent improvement.