心音身份识别技术的研究
摘要
This paper presents the basic principle and experimental method of identity recognition based on heart sounds. Firstly, the characteristics of heart sounds and feasibility of Heart sound as a biometric are analyzed. Then, the synthetic model of heart sounds is constructed based on wavelet family, and the characteristics of the two heart sounds are compared by the phase diagram of the feature vector, we make use of cepstral subtraction to eliminate the effects caused by the type difference and probe position change of the stethoscope. Finally, the characteristic parameters of heart sounds are extracted by using the liner band of the frequency cepstral of heart sounds (HS-LBFC) with a specified configuration, and the similarity distance is adopted for heart sounds identification. To highlight the differences between various heart sound signals in the time domain or the frequency domain, we make further research on the construction method of the heart sound wavelet, the calculation method of the parameters of the synthetic model, the selection of the characteristic parameters of heart sound and the corresponding data processing technology. The experimental results show that above method exhibits nice performance and practicality.