Study On Heart Sound Identification

Along with the rapid development of social economy and the rapid popularization of networkcommunication technology, the requirements of information’s security, accuracy, privacy isincreasingly important, so more attention are putting forwards on looking for more effectiveauthentication and identification techniques. Facing this demands and challenges， Biometrictechnology is developed rapidly. Compared to the traditional identification methods (such as keys,badges, personal documents, ID cards and other authentication–based cryptography), biometricsidentify human using inherent physiological characteristics and behavioral characteristics, so theywill not be forgotten/lost, etc. with strong security and reliability. Recently, the biometrictechnology which have been commercialized include fingerprint recognition, voice recognition, irisrecognition and so on. Heart sound is one of the important human physiological signal withbiometric which is universality, uniqueness, and can be collected in nature, At the same time，heartsound is endogenous signal，can not be imitated or copied easily.This paper mainly studies heart sound signal authentication algorithm. Firstly, the paperanalyzes the generation mechanism and the time-frequency domain characteristics of the heartsound signal, builds a variety of heart sound database, preprocesses the heart sound signal(denoising, framimg, window). And then this paper proposes a identification method based on heartsound signal spectrum of THT and vector quantization VQ, designs various types of experiments totest the performance of heart sound identification algorithm, obtains a higher recognition rate, thispaper’s mainly works are as follows:1、The paper describes the common biometric technology, introduces the background of heartsound signal as biometric, and a brief description of the generation mechanism of heart soundsignals and time-frequency characteristics.2、Proposes the heart sound signal preprocessing algorithm based on Ensemble EmpiricalMode Decomposition (EEMD). Depending on the different distribution characteristics of heartsound signal and noise, EEMD decomposes the heart sound signal, gets the energy distributioncharacteristics figure of Intrinsic Mode Function, according to confidence curves of the noise toselect the retention IMF, chooses threshold for every retained IMF to denoise. Then reconstructsthe heart sound signal; According to the quasi-periodic of heart sound signals, detects the signalquality and improves the system robustness.3、Proposes the individual identification algorithm based on THT (Teager Huang Transform) boundaries spectrum and vector quantization. Firstly, windows and cuts off the denoised heartsound signal; then does the THT transform for each heart sound, extracts the boundary spectralfeatures; achieves personal identification using Vector Quantization.4、Designs series of experiments to test the performance of the algorithm. Analyzes theinfluence of different parameters for recognition performance, gets the optimal analysis parameters.Studies the algorithm performance of types of heart sound database (different test locations,different moods and different states). The results show that the proposed algorithm achieved abetter recognition rate.This paper provides a heart sound signal authentication method with a high correct recognitionrate, which laid the foundation for further development of heart sound identification system.