The Research On PPG Signals Reconstruction Based On Compressed Sensing

With the development of embedded technology, sensor technology and wireless communication technology, medical equipment towards a more portable, intelligent,wireless and network direction. A variety of portable and wearable medical equipment is widely used in human health and disease diagnosis. But for these kinds of equipment, if the traditional Nyquist sampling theorem is used for signal collection,not only it will occupy a lot of storage space, but the data storage and transmission process will take a lot of system processing time,wasting system resources.Compressed sensing technology is a novel signal processing technology, which is used to compressed sampling of signal and reduces the costs of data collection,storage and transmission. Photoplethysmography(PPG) signals contain abundant physiological and pathological informations, are commonly used in clinical diagnosis and medical care equipments. Resarching CS technology of PPG signals is important for remote monitoring and disease diagnosis.The sparsity of PPG signals, construction of measurement matrix and reconstruction were researched in this thesis.Adaptive compressed sensing method is proposed according to the character of PPG signals.The main contents are as follows:(1) Due to PPG signals aren’t sparse in time domain, the sparsity of the PPG signals is researched in the DCT, DFT and DST transform domains. Experiments show that the PPG signals have good sparsity and is best in the DCT domain. The reconstruction performances of measurement matrix and several common algorithms are analyzed and compared under different compression ratios.(2) Some reconstruction algorithms cann’t accurately estimate the sparse degree,reconstruction precision isn’t hight and real-time performance is lower. Based on reconstruction method of sparse adaptive matching pursuit(SAMP), the algorithm of modified adaptive matching pursuit(MAMP) was proposed by the sparsity of the PPG signals. The proposed method is used in the PPG signals of simulation and actual acquisition. The experiment results show that algorithm could accurately and quickly estimate the signal sparsity, and had good anti-noise performance; the reconstruction speed of the algorithm was faster, the error was lower, and is small by noise and sampling frequency than algorithms of SAMP and orthogonal matching pursuit(OMP).It is employed to reconstruct PPG signals.(3) According to the characteristics of the reconstructed before and after signal may be different, pulse rate variability signals are extracted from the PPG signals and further analysed the accuracy and real-time performance of the proposed method from the extracted features.The some characteristic parameters extracted are used to classify normal people and patients with coronary heart disease(CHD) by some machine leaning algorithms, the classification results have a high accuracy.