Biometric Identification Method For Heart Sound Based On Periodic Segmentation

Heart sound signals are natural physiological signals generated by the heart through the opening and closing of the atrioventricular and semilunar valves and have great advantages in the field of biometrics.Therefore,based on the characteristics of heart sound signals,this thesis proposes two heart sound biometric methods based on the periodic segmentation algorithm.(1)a new method of biometric characterization of heart sounds based on periodic segmentation and multimodal multiscale dispersion entropy is proposed.Firstly,the heart sound is periodically segmented,and then each single-cycle heart sound is decomposed into a group of intrinsic mode functions(IMFs)by improved complete ensemble empirical mode decomposition with adaptive noise(ICEEMDAN).These IMFs are then segmented to a series of frames,which is used to calculate the refine composite multiscale dispersion entropy(RCMDE)as the characteristic representation of heart sound.In the simulation experiments,carried out on the open heart sounds database Michigan,Washington and Littman,the feature representation method was combined with the heart sound segmentation method based on logistic regression(LR)and hidden semi-Markov models(HSMM),and feature selection was performed through the Fisher ratio(FR).Finally,the Euclidean distance(ED)and the close principle are used for matching and identification,and the recognition accuracy rate was 96.08%.Experimental results show that the method is effective for heart sound biometrics including pathological heart sounds.However,on the one hand,the multi-modal and multi-scale dispersion entropy method is only a measure of the complexity in the time domain of a single-cycle heart sound,and may not truly reflect all the characteristic information in the single-cycle heart sound.On the other hand,the above studies are all conducted on the open pathological heart sound database,that is,only the influence of pathological heart sounds on the heart sound biometric recognition algorithm is considered,and other factors are not considered.Therefore,another heart sound biometric recognition method is proposed for the above problems.(2)A cepstrum coefficient of heart sound frequency based on periodic segmentation is proposed for heart sound biometric identification.This method first uses the heart sound segmentation algorithm to divide the input heart sound into a series of single-cycle heart sounds;then divides the single-cycle heart sounds into a series of overlapping segments,where each small segment represents a heart sound frame;then calculate the heart sound frequency cepstrum coefficient of each heart sound frame,combine the heart sound frequency cepstrum coefficient of each frame in the same cycle into a feature,and finally,the single-cycle heart sound frequency cepstrum coefficients(HFCC)are used as the character representation of heart sound.Simulation experiments were performed on an open heart sound database HSCT11 containing healthy subjects of different genders.The feature representation method was combined with the heart sound segmentation method based on LR-HSMM,and the ED and the close principle are used for matching and identification,the recognition rate reached the highest at 94.56%.The experimental results show that the method is effective for biometric recognition including heart sounds from subjects of different genders.To improve the practical application value of the above two methods,this thesis conducted two additional application experiments.In the first experiment,the method(1)was applied to the 80 heart sound database constructed by the heart sounds of 40 healthy volunteers from different age groups.To investigate the effect of single-cycle heart sounds with different starting positions on the performance of the algorithm,research shows that the single-cycle heart sounds from the first heart sound to the start of the next first heart sound have the highest recognition rate of 97.50%.Experiment two applies the multi-modal and multi-scale entropy proposed by method(1),the cepstrum coefficients of the heart sound frequency proposed by method(2),and the fusion characteristics of them to the combined heart sound database,which is composed of the self-built heart sound library and the public heart sound database of Michigan,HSCT11.The research shows that the method proposed in this thesis is less affected by these three factors,and the recognition rate of the fusion feature is the highest,which is 97.84%.