Extraction Of Respiratory Signals And Respiratory Rates From The Photoplethysmogram

Respiratory rate(RR)is one of the important indicators for hospitals to prevent sudden ab-normalities in patients.In a study on abnormal breathing,54% of cardiac arrest patients had at least one RR >27 bpm three days before the cardiac arrest,so it is very important to mon-itor the patient's respiratory rate.With the advancement of photoplethysmography(PPG),PPG is widely used in respiratory monitoring.PPG detection is non-invasive and reliable,and lower cost,which has a positive effect on clinical application and daily monitoring.The primary task of detecting the respiratory rate from the PPG signal is to extract the respiratory modulation signal,and then perform spectrum analysis on it.Most of the algorithm accu-racy rates of existing analysis methods do not meet clinical requirements,and the algorithm performance is difficult to be further improved.On account of the above problems,proposed the AC-AR algorithm based on the analysis and summary of the existing related research,which aims to improve the accuracy of the respiratory rate extraction algorithm from the algorithm level.It has good practical value for clinical detection of respiratory rate.With the open source data set,the accuracy of the AC-AR method are verified by comparison with other algorithms.The specific research content and research contributions of this thesis are as follows:(1)Improved peak detection algorithm and artifact detection algorithm.Using the Incremental-Merge Segmentation algorithm to extract the morphological features of the signal from the time domain,and using an adaptive threshold to divide the PPG signal into two kinds of segments,impulse and artifact.Then,the improved peak detection method is used to obtain the peak value,and three types of respiratory modulation signals(amplitude modulation,fre-quency modulation,and baseline drift)are calculated.At the same time,the dicrotic wave is detected to reduce the peak error,and the signal quality index is introduced to evaluate the signal quality.(2)Proposed a method of autocorrelation analysis combined with autoregressive model to analyze the frequency spectrum of respiratory modulation signal,and compare it with short-time Fourier transform,autocorrelation analysis,and autoregressive model respectively.At the same time,since the regularity of the autocorrelation signal can indirectly reflect the qual-ity of the respiratory modulation signal,a fusion method based on the quality of the autocor-relation signal as the selection criterion is proposed.Using open source data sets for testing,the algorithm accuracy rate is 96.38%,and the average absolute error is 0.12±0.36 bpm.On the one hand,the research in this thesis reduces the computational complexity of peak detection and artifact detection,and meets the clinical requirements for the performance of the respiratory rate extraction algorithm.On the other hand,it combines autocorrelation analysis and autoregressive model to analyze respiratory modulation signal spectrum,which improves the accuracy of the respiratory rate extraction algorithm.The research in this thesis is a beneficial exploration of the algorithm for extracting the respiratory signal and respiratory rate from the PPG signal.