The Study Of Heart Sound Analysis Based On Nonlinear Fractal Theory

The heart sound signal is one of the most important human physiological signals,which contains much physiological and pathological information of the human cardiacsystem. Because the heart is a complex nonlinear dynamic system, the heart soundsignals are nonlinear signals and inherent complexity. The analysis method is built onthe heart sound signal linear model based on the fractal theory as an important branch ofnonlinear science, and it can well reveal the nonlinear process with special regularity.This paper applies fractal theory to the analysis of heart sound signals from a nonlinearpoint of view. Also box-dimension algorithm will be used to the quantitative analysis ofthe heart sound signals. It provides a noninvasive method and a valuable reference forassessing cardiac function.Firstly, the paper analyzed the characteristics of heart sound signals respectivelyfrom time-domain, frequency-domain and time-frequency domain. The STFT (ShortTime Fourier Transform) and Welch method were used to estimate the power spectrumof heart sound and the murmurs. Results showed that the frequency range of normalheart sound and that of murmur components were different, and the heart sound signalsin time domain and frequency domain characteristics had certain regularity. It laid thefoundation of the subsequent analysis of heart sound signals.The pretreatment of heart sound signal was wavelet threshold de-noising. Based onthe reconstructed factor the research selected the suitable wavelet function and the resultindicated that coif3wavelet had a good de-noising for heart sound signals. Appliedfractal theory to heart sound signals, the premise need to ensure the heart sound signalshad fractal characteristic. In this paper, fractional Brownian motion model and the heartsound signal model were for comparative studies, qualitative proved the heart soundsignals with self-similarity and scale-free nature two important fractal characteristics.That provided the basis for the subsequent use of the fractal theory of heart soundsignals. Then the paper gave a variety fractal dimension of the principles and methods,such as Box-counting dimension, correlation dimension, information dimension,Hausdorff dimension and similarity dimension. The final choice on the research of heartsound signals was Box-counting dimension.Finally, the heart sound signals of three sample populations were calculated thoughbox dimension algorithm. Subjects concluded153healthy pregnant women and64 pathological pregnant women,153healthy pregnant women and161healthynon-pregnant women,97athletes and65college students. Results showed that the FDvalue of healthy pregnant women (1.27±0.29) was lower than the FD value of healthynon-pregnant women (1.31±0.41). Also, the FD value of healthy pregnant women(1.27±0.29) was lower than the FD value of pathological pregnant women (1.38±0.05).In the third group, the post-exercise FD value of athletes (1.30±0.065) was lower thanthe FD value of resting state (1.39±0.037). And the post-exercise FD value of collegestudents (1.32±0.054) was lower than the FD value of resting state (1.41±0.065). All theresults had differences in statistical significance. The results showed that the fractaldimensions were changed under stress with a significant difference. Also the researchmade a comparative analysis with other index, such as D/S, S1/S2and HR, for assessingcardiac function.Results showed that the box-dimension value can quantitatively describe thecomplexity of the heart sound signals, and different groups of heart sound signals withdifferent fractal dimension values. This study allows doctors and researchers tounderstand the changes of cardiac function under stress. And it provides more valuableinformation for clinical and more basis for detailed heart protection.