Study On Feature Extraction And Classification Recognition Of Photoplethysmography Signals

Photoplethysmography(PPG)signal is a kind of physiological signal by means of optoelectronic sensing technology detecting the change of blood volume in human tissues.It has great research value and broad application prospect in the detection of physiological parameters,in the development of medical equipment,in the study of cardiovascular diseases and pulse diagnosis objective.For the research of feature extraction and classification recognition of PPG signal,the following works have been done in this paper:Firstly,on the basis of clearing the mechanism of PPG signal production,waveform and signal features were introduced,the principle of transmission and reflection photoelectric detection was studied,and the influence of incident light wavelength on the detection results and the tissue structure and blood distribution of the fingertip was also researched,which made theoretical basis and practical significance clearer to detect PPG signal in human fingertip based on sphygmography theory.Secondly,in view of the analysis of the characteristics of noise and interference which affected the quality of PPG signal,photoelectric sensing and signal conditioning circuits were designed.Then,the conditioning circuit was simulated using Multisim 14.0 software,and the simulation results showed the conditioning circuit made voltage signal amplified and noise was filtered well.At the same time,the voltage rise circuit was designed to meet the input requirements of A/D conversion chip and A/D conversion circuit was designed as well.Lastly,adopting transmission detection technology and choosing the 650 nm laser source,we established the experimental platform.When the output port of the conditioning circuit was connected to the oscilloscope,the PPG signal from the fingertip could be extracted.Thirdly,to obtain high-quality PPG signal,PPG signal was preprocessed by means of the method of wavelet transform and the HHT method based on EEMD.The two methods both got higher signal-to-noise ratio and lower mean square error,resulting in a good denoising result.Meanwhile,the analysis of time domain was made by taking the differential threshold method,pulse area method and waveform fitting method.Power spectrum analysis and cepstrum analysis was adopted to do frequency domain analysis.And short time Fourier transform,wavelet transform and HHT methods were taken to do time-frequency mixing domain analysis for PPG signal.At last,these methods could well extract feature parameters of PPG signal.Fourthly,to distinguish slippery,taut and astringent pulse,the differences between three kinds of pulse in time domain,frequency domain and time-frequency mixing domain were introduced based on waveform fitting,power spectrum analysis and wavelet analysis.Then,the pulse signal was decomposed by five layers of wavelet,and the ratio of energy distribution in the first four bands was extracted to be compared with the standard slippery,taut and astringent pulse data to judge the pulse category.