Research Of Multi-Position Fusion Heart Sound Identification System

The importance of effective individual identity is more and more prominent in a highly informationization of modern society. Traditional identification ways are liable to be lost or faked, which are unable to meet the needs of modern society on security. Biometric identification echnology, as an approach of using body physiological or behavioral characteristics to realize identification with higher safety and reliability, has gradually played an important role in different fields. Biometrics has the advantage of higher security and convenience, but it faces many challenges in practical application, especially the problem of safety. Therefore, exploring new biometric identification technology of high security is the hot spot of research in the world. Heart sound can well meet the requirements of security and reliability in the field of identification since it has unique properties such as uniqueness and unforgeability. At present, heart sound identification is getting the wide attention of scholars and becoming one of the research hotspots in the field of identification.Based on the related research of pre-processing, feature extraction and pattern matching, this paper studies the differences of heart sounds at different acquisition positions and proposes multiple position fusion method to eliminate the problems that may occur in the recognition process in which different acquisition positions lead to the decline of the recognition rate. The multi-position fusion heart sound identification system is implemented in the Android smart phone. The researches of this paper are mainly as follows:(1) In view of the pre-processing of heart sound signal, this paper proposes heart sounds automatic segmentation algorithm. Based on the quasi-periodic characteristics of heart sound, the segmentation algorithm extracts heart sounds envelope and detects peak points and boundary points to locate each component of heart sound. The heart sound automatic segmentation algorithm is tested on both normal heart sound and complicated heart sound, the experimental results show that each component of heart sounds is located precisely, which proves that heart sounds automatic segmentation algorithm has strong robustness and accuracy.(2) This paper proposes a new heart sounds identification method based on Bark Frequency Cepstrum Coefficients (BFCC) and Weighted Vector Quantization (WVQ). This new method first analyzes the basic principle of BFCC and parameter characteristics and then proposes heart sounds identification method combining BFCC+ABFCC and WVQ. The optimal parameters of proposed method are obtained from experiments.(3) The differences of heart sounds at different acquisition positions are studied and this paper proposes feature level fusion method to realize multi-position fusion of heart sounds. The experiments confirm that position fusion can effectively reduce the influence of the collecting position sensitivity of heart sound recognition and eliminate the identify differences caused by the different acquisition location.(4) The heart sound identification application based on Android operating system is developed, which mainly consists of four functional modules:Bluetooth connection, user registration, user identification and system management. This application can receive and display heart sounds, process and extract heart sound feature, realize position fusion and code training and so on.This paper provides a new heart sounds position fusion identification system based on Android smart phone, which solves the heart sounds position sensitive issue in the process of identification and provides a safe, reliable and convenient new technology.