Research And Application Of Multi-channel Cardiopulmonary Sound Separation Method

Cardiopulmonary sound auscultation is one of the important diagnostic methods for the health of the human cardiopulmonary system.The traditional auscultation process must be completed by professional doctors,which is highly professional and less convenient,and is difficult to be popularized in daily life.The emergence of single channel electronic stethoscope greatly alleviates this phenomenon.The single channel electronic stethoscope can well collect the original cardiopulmonary sound data and give the diagnosis results through detection algorithms or remote interrogation.However,the single channel electronic stethoscope has poor anti-interference ability and low collection efficiency,and is prone to be affected by the collection environment,resulting in data distortion and misdiagnosis.In addition,heart sounds and lung sounds are physiological signals that interfere with each other,and independent heart sounds and lung sounds can help doctors locate abnormal cardiopulmonary sounds more accurately.Therefore,how to separate cardiopulmonary sounds is a research hotspot at present.In this thesis,a multi-channel electronic stethoscope connected by wireless network is designed.The multi-channel electronic stethoscope is composed of front-end acquisition circuit,acquisition channel expansion circuit,analog digital conversion circuit,main control circuit,data storage circuit and network transmission circuit.It has the characteristics of strong anti-interference ability and high efficiency of cardiopulmonary sound acquisition.Based on the self-designed multi-channel electronic stethoscope,a systematic processing method suitable for multi-channel cardiopulmonary sound separation is constructed.The method is divided into two main stages: cardiopulmonary sound noise reduction and cardiopulmonary sound separation.In the noise reduction stage,the advantage of multi-channel acquisition is utilized,combined with Wiener filtering and spectral subtraction to filter out environmental noise in the cardiopulmonary sound.Then,the residual circuit noise in the cardiopulmonary sound is suppressed using the method of Complete Ensemble Empirical Mode Decomposition with Adaptive Noise combined with a dual threshold function to improve the signal-to-noise ratio of the cardiopulmonary sound data.In the separation stage,the multi-channel cardiopulmonary sound data is preprocessed using a multi-channel cardiopulmonary sound data merging method to preserve high quality signals,and then combined with Independent Vector Analysis to separate cardiopulmonary sounds.After comparison and verification,the stability and separated cardiopulmonary sounds data quality of the cardiopulmonary sound separation method in this thesis are better than those cardiopulmonary sound separation method based on Empirical Mode Decomposition combined with Fast Independent Component Analysis and those cardiopulmonary sound separation method based on Nonnegative Matrix Factorization Recursive Sparse Representation.Finally,the supporting software of multi-channel electronic stethoscope is designed to process the collected multi-channel cardiopulmonary sound data,realize the functions of cardiopulmonary sound separation,waveform display,data storage,etc.,which provides convenience for remote auscultation and non-contact auscultation,and has certain practical value.