| Biometric identification is widely used to identify identity information in the world.Compared with passwords and certificates,biometric identification is easy to crack or copy.It has high safety and good research and application environment.Heart sound signal is one of the biological characteristics of human body.Heart sound signal in motion and quiet states includes the heart sound signal collected in quiet and moving state.The existing researches are usually conducted from the perspective of quiet and moving heart sounds,and the pattern recognition of heart sounds in motion and quiet states is a difficult problem.This thesis studies the signals of heart sounds in motion and quiet states from the overall perspective,and mainly completes the following four aspects of work: solve the problem of heart sounds acquisition in motion and quiet states,in-depth analysis of the characteristics of heart sounds in motion and quiet states,put forward a simulation model of heart control conversion function,and study the problem of pattern recognition of heart sounds in motion and quiet states.The main contributions and innovations of this thesis are as follows:(1)The acquisition of heart sounds in motion and quiet states was studied.This thesis introduces the single-channel and multi-channel heart sound signal acquisition equipment designed and made by the author.This thesis analyzes the collection method of heart sounds,and provides the hardware foundation for the collection of heart sounds in quiet state and movement state.Furthermore,the location of heart sound acquisition was studied.By analyzing the characteristics of gender difference,acquisition stability difference and the mechanism of heart sound production,the pulmonary valve auscultation area was determined as the optimal location of single channel heart sound signal acquisition.(2)The characteristics of heart sounds in motion and quiet states were studied.The characterization methods of parallel and serial features of heart sound signal based on multi-temporal window were proposed,and the multi-feature characterization methods of heart sound signal based on K-entropy feature analysis,RCMDE feature analysis and RQA feature analysis were introduced.In particular,RCMDE feature analysis method and RQA feature analysis method were used to analyze moving heart sound.On this basis,the heart sound signals were compared and analyzed by the sound direction vector,motion response curve of heart sound,trend diagram of state change,difference value of state change,quantitative recursive graph,cross quantitative recursive graph,RQA analysis,CRQA analysis and chaos characteristics.These characteristics can effectively represent the individuality of the silent heart sounds and the moving heart sounds,and also the commonness of the stationary heart sounds and the moving heart sounds.The results suggest that in the absence of cardiovascular disease,exercise generally has only a quantitative,not a qualitative,effect on heart sound signals.(3)A transformation simulation model of heart sound control conversion function was studied.The heart can use its own control switching function to gradually switch the exercise state heart sound to the quiet state heart sound.Based on the chaotic characteristics of heart sound signal,a transformation simulation model of heart control conversion function is constructed in this thesis.With this model,the motor heart sound signal can be regulated by the stationary heart sound signal,so that the heart sound in motion state can be approximated to the heart sound in quiet state,and the control conversion function of the heart can be simulated.The results of the simulation model show that the conversion of the heart sounds in motion and quiet states must have homology,and heart sound in motion state can be converted to heart sound in motion state for analysis under a unified framework,which is conducive to understanding and analyzing the heart sound signals in motion and quiet states from the height of unified theory.(4)The pattern recognition of heart sounds in motion and quiet states was studied.According to the characteristics of heart sound signals in motion and quiet states,a pattern recognition network(HEINet)with high recognition rate and high reliability is proposed,and its design requirements and principles are discussed.According to the design method of HEINet,HEINet for heart sound signals in motion and quiet states,HEINet for HSCT11 heart sound database data and HEINet for MIT-BIH arrhythmia database are designed respectively.The design process has the characteristics of simplicity,speed and convenience.The results show that the recognition rate of our own heart sound database in motion and quiet states can reach 99.10%,the recognition rate of HSCT11 heart sound database and MIT-BIH arrhythmia database are 99.56% and 96.37%,respectively.Therefore,the design method of HEINet has obvious universality to heart sound and electrocardiogram,two physiological signals from the heart.Pattern recognition of heart sounds in motion and quiet states based on multidisciplinary cross fusion of academic thought,the theoretical simulation,biological experiment and the method of combining the electronic technology and the technology of multi-dimensional path,to carry out the multifaceted heart sound signals in motion and quiet states modeling and experimental research,explore the mixing of the change of heart sounds,the basic mechanism of acoustic and signal transfer law,the mix of heart sound pattern recognition technology.This not only expands the research content of feature extraction,characterization and application based on heart sound,but also provides theoretical and experimental basis for making full use of heart vital information and manufacturing new heart sound recognition equipment. |
PDF Full Text Request