Analysis And Research On Heart Sound Signals Based On Compressed Sensing

In recent years,in order to make heart disease monitoring easier and faster,recent work utilize the new developed technologies of the Internet of Things(IoT),and wearable sensor,which can provide a new way to monitor the heart disease.The way analyzes the development trend of heart failure and predicts of sudden death through the monitoring of heart sound(HS)signals to achieve the purpose of auxiliary diagnosis.Especially for long-term synchronous acquisition of HS signals generated by the aortic auscultation area,pulmonary auscultation area,mitral auscultation area,and tricuspid auscultation area,the amount of data generated become greater.When the generated HS signals are synchronously collected for a long period,the data is large.Therefore,some related problems are waiting to solution,such as how to store HS signal data,and how to make diagnostic equipment stand by for a long time to ensure it stable operation.Compressive Sensing(CS)is a new theoretical framework for the acquisition and reconstruction signal from a small number of sampling data.It can sample the signal at a lower frequency and reconstructs the signal with high probability.This provides a possible solution for wearable devices of long time running that store a large amounts of HS signal data.This thesis proposes a parallel compressive sensing model for multi-channel synchronous acquisition of HS signals.This model provides an effective way to solve the problem of a large amount of data storage generated by wearable multi-channel HS signals device,and can ensure that the wearable devices can run for a long time in an energy-saving state.According to sparse characteristics of multi-channel HS signals and compressive sensing related theory,this thesis proposes a parallel compressive sensing model for multi-channel synchronous acquisition of HS signals.This model can synchronously collects multi-channel HS signals and compresses them in parallel and reconstructs them,and then it can classify and identify abnormal HS signals by their feature extraction.This model provides an effective way to solve the problem of a large amount of data storage generated by wearable multi-channel HS signals device,and can ensure that the wearable devices can run for a long time in an energy-saving state.In addition,it can be applied to parallel compressive sensing processing to other multi-channel synchronous acquisition biological signals.The model proposed in this thesis has five main characteristics: A.Maximum compression ratio to achieve the maximum compression and preservation of multi-channel synchronous acquisition signal data in order to reduce the storage space;B.Optimal reconstruction,this thesis proposes using least squares interior point method as reconstruction algorithm,in order to ensure the recovery of compressed sensing reconstructed signal and make the reconstructed signal and original multi-channel as far as possible.Synchronized acquisition signals have high similarity;C.Minimum cost,reduce the calculation cost of software and hardware when reconstructing multi-channel signals;D.Aiming at specific multi-channel synchronous acquisition signals,parallel compression sensing of multi-channel acquisition signals is realized based on the characteristics of time domain,frequency domain and sparse characteristics of multi-channel signals;E.Minimum storage capacity,compression of multi-channel acquisition signals by single-channel serial method.Signal is still the compressed data quantity for storing multi-channel signals,which fails to achieve the goal of minimum storage.Choosing the compressed sensing of the main signal only needs to store the compressed data of the main signal.Secondly,in order to meet the needs of identity recognition,thousands of people's physiological signals are stored.In order to analyze a person's activity pattern and health status,and to store physiological signals selectively for a long time continuously,this thesis combines physiological signal processing technology with compressed sensing technology,and proposes a classification of non-reconstructed compressed sensing for physiological signals.Method.This method can classify and store the heart sound signals in the compressed domain,and has a high classification and storage accuracy.Finally,the parallel compressed sensing model of multi-channel synchronous acquisition signals is used to process and analyze four normal and four abnormal heart sounds.From the experimental results,the model has certain reference value for monitoring and storing heart sounds for a long time in the future,and for analyzing and predicting sudden death of heart failure.