Research On Noise Reduction And Recognition Algorithm Of Heart Sound Signal

Heart sound signals contain a lot of information about cardiovascular diseases,and auscultation of heart sounds is an important means of diagnosing cardiovascular diseases.However,due to the interference of various factors such as the environment and the heart,it has increased the difficulty of auscultation of heart sounds.The traditional stethoscope diagnosis method only relies on the doctor’s auscultation experience for diagnosis,resulting in a low diagnosis accuracy.This paper studies the automatic classification and recognition of heart sound signals through signal processing and pattern recognition,which provides an objective basis for the diagnosis of cardiovascular diseases and improves the diagnosis efficiency of cardiovascular diseases.The main research work is as follows:(1)Noise reduction processing of heart sound signal.Considering that the collected heart sound signal contains noise,a combined complete ensemble empirical mode decomposition(CEEMD)and a heart sound noise reduction algorithm with improved wavelet threshold are proposed.The heart sound signal is decomposed into intrinsic modal function(IMF)of different frequencies by using CEEMD,and the high-frequency IMF component are selected to perform noise reduction processing using the improved wavelet threshold algorithm in this paper.The denoised IMF component and the low-frequency IMF component are reconstructed to obtain the denoised heart sound signal.Under different noise intensities,the combined CEEMD and improved wavelet threshold algorithm,CEEMD algorithm and wavelet threshold algorithm are used to denoise the heart sound signal respectively.The simulation results show that the noise reduction algorithm in this paper can not only improve the signal-to-noise ratio of heart sound,but also reduce the root mean square error.Among the three algorithms,the noise reduction effect is the best.(2)Feature extraction of heart sound signal.In view of the non-stationary characteristics of the heart sound signal,in order to display the characteristics of the heart sound signal more comprehensively,the characteristic parameters of the heart sound signals are extracted and analyzed from three aspects: time domain,frequency domain and cepstral domain.The short-term energy,short-term zero-crossing rate and short-term autocorrelation function of the heart sound signal are extracted as time-domain feature parameters;the frequency-domain feature parameters of the heart sound signal are extracted by short-time Fourier transform;The 12-dimensional mel cepstrum coefficients are extracted by mel frequency cepstrum operation as the characteristic parameter of cepstrum domain.Save the16-dimensional feature parameters extracted from the time domain,frequency domain,and cepstral domain for subsequent heart sound recognition.(3)Classification and recognition of heart sound signals.Aiming at the problem of poor parameter setting of traditional support vector machines(SVM),which leads to low recognition accuracy,an improved artificial fish swarm algorithm(IAFSA)is proposed to optimize the heart sound recognition method of SVM(IAFSA-SVM).First,the adaptive field of view radius and moving step size are introduced,and the IAFSA algorithm is proposed and tested.The simulation results show that,compared with the artificial fish swarm algorithm(AFSA),the convergence and globality of IAFSA are improved.Then,the IAFSA algorithm is used to optimize the SVM parameters,and the 16-dimensional heart sound features are input into the IAFSA-SVM for identification,and the identification sensitivity is 98.44%,the specificity is 95.59%,and the accuracy is 96.50%.