A Research On Heart Sound Graphical Processing Methods Based On Heart Sounds Window Function

Currently, the one-dimensional signal processing method for heart sound analysis and recognition is the mainstream in researches. In order to get more intuitive features manifestation, to improve the effect of classification, to expand the heart-sound recognition field, and to achieve heart-sound on-line identification, this paper puts forward a heart-sound texture feature extraction and recognition algorithm, which is based on heart-sound window function and the combination of heart-sound and image processing technology. Heart-sound texture map records the heart-sound signals intuitively and truly, and contain a large number of individual physiological information, such as identity information and health status, etc.Firstly, heart-sound composition model is given, heart sound time-frequency diagram and heart-sound texture map are defined, how to utilize heart-sound window function and short-time Fourier transform to obtain two-dimensional heart sound time-frequency diagram is discussed. After that, in the light of the characteristics of heart-sound, the structure principle and implementation method of heart-sound window function are mainly studied. The coefficient of heart-sound window function is adjusted by the different features for normal heart-sound and pathological heart-sound, to meet the heart sound window function requirements mentioned in the structure of the principle. Finally, the normal and pathological heart-sound texture feature extraction and identification are separately realized by the improved pulse coupled neural network model(IPCNN) and the Heart sound texture figure angular point correlation fast recognition algorithm.Simulation experiments show that compared with the traditional window function, the heart- sound time-frequency diagram obtained by heart-sound window function has a clearer and noise well suppressed texture, the adaptive method can effectively identify different coefficient in the heart-sound window function. Furthermore, compared with other three kinds of typical recognition methods, IPCNN has the lower computational cost and higher recognition rate. Especially heart-sound identification has high recognition rate and high speed by using heart-sound texture figure angular point correlation fast recognition algorithm, which can realize the on-line identification. So, we can arrive at the conclusion that the method for heart-sound feature extraction and recognition based on image processing techniques is the effective one.