ESTIMATION OF THE COEFCIENTS OF THE AR MODEL AND BISPECTRUMOF LUNG SOUNDS ADOPTING HIGHER-ORDER CUMULANTS
基于高阶累计量的肺音信号AR模型参数和双谱估计

Based on the fact that the signals of lung sounds behave a non-Gaussian characteristics, non-Gaussianility AR model for lung sounds is set up, and higher-order cumulants technique is used to investigate the non-Gaussianility properties of lung sounds, estimate the coefficients of the AR model and bispectrum of lung sounds for the purpose of overcoming the drawback of conventional spectrum methods. Experiments have shown that sources lung sounds consist of random non-Gaussian white noise sequence, periodic impulses, and intermittent impulses. The estimated filter response has a low pass characteristics. Bispectrum analysis can reveal extra information and have an insight into the origin and the transmission of lung sounds.