The Pre-processing Of PPG Signal And Extraction Method Of Dynamic Spectrum

The noninvasive measurement of human blood components is one of the frontiers of biomedical engineering.Optical method has been a hot spot for noninvasive measurement methods in recent years.The most representative method is near-infrared spectroscopy among them.It has good prospects because it can be done with high accuracy,high speed and low cost.However,measurement conditions and individual differences of the measured subjects have always been major limitation for useing into the clinical.Thereupon,dynamic spectrum(DS)based on PPG signal can eliminate them effectively and has been widely concerned.According to the present research status of noninvasive measurement of blood components,in this paper,the theory about eliminating measurement conditions and individual differences of the measured subjects using DS was briefly introduced at first.On the basis of dynamic spectrum method,the following aspects were mainly discussed.A dynamic spectral data acquisition system with LED as the compensating light source was designed,and the spectral data of the subjects were collected by the system.The basic characteristics of PPG signal and the importance of PPG signal preprocessing in dynamic spectrum were analyzed.In order to improve the effectiveness of PPG signal pre-processing,a method of combining double tree complex wavelet transform(DTCWT)with threshold analysis and mathematical morphology was proposed.This method can improve the problem that the different noise can not be removed at the same time in PPG signal preprocessing,and improve the signal to noise ratio(SNR)of PPG signal.Based on the principle and characteristics of dynamic spectrum,this paper further proposes a DS extraction method based on signal source,that is,independent component analysis(ICA)combined with DTCWT.In order to verify the effectiveness of the method,the DS data of 151 subjects data collected by spectral system were extracted by the new method.And their blood component concentration modeling were carried out.The method was compared with the currently used frequency domain analysis and single trial estimation.The DS data extracted from the subjects were modeled by partial least squares(PLS).The accuracy of the prediction results of the frequency domain analysis and the single trial estimation was 63.71% and 78.83% respectively.And the accuracy of ICA combined with DTCWT was 90.72%.The high accuracy prediction results verified the feasibility and effectiveness of ICA combined with DTCWT.This study extracts the dynamic spectrum data from signal source instead of the traditional time domain and frequency domain,which improves the accuracy of the dynamic spectrum method to predict the blood components and broaden the ideas of dynamic spectrum study.