Research Of Blood Component Content Detection Method Based On Dynamic Spectrum

Invasive detection of blood composition is the main clinical detection methods currently,with the disadvantages of high cost,slow detection and produce a large number of one-time consumables and so on.So the noninvasive detection method gets more and more attention of many scholars.But due to the difference between individuals and background organization interference,the detection accuracy can not meet the requirements of clinical application.Dynamic spectrum detection method of photoelectric volume pulse wave breaks this limitation.It can eliminate organization interference and the differences between individuals on principle.There are various noises of different frequency in photoelectric volume pulse wave,so eliminating the noise of signals effectively can improve the accuracy of noninvasive blood component detection.In addition,the extraction of dynamic spectrum value and forecast model building also plays a key role for the testing results.This article introduces the basic principle of dynamic spectrum as well as the specific detection process firstly;Then introduces the methods used in each stage.In the pretreatment stage,it removes the baseline drift using compensation method,and then removes the high-frequency random noise using the adjustment of spatial correlation filters.In the extraction stage of dynamic spectrum data,this paper proposes a new extraction method.It makes up for the deficiency of the extraction method of frequency domain,and overcomes the phenomenon of spectral overlap.In the model building stage,each sample consists of 606 wavelength dynamic spectral values,so first to take dimension process for all the dynamic spectrum value samples with the method of principal component analysis.And finally choose the first 39 principal components.It's known that BP neural network has good ability to deal with nonlinear problem,but it also has the limitations of global optimization.So using the genetic algorithm and BP neural network to complement each other.The final results prove that this method can effectively improve the accuracy of noninvasive detection of the blood's components.