Research Of Technology On The Phonocardiogram Feature Extraction Based On Image Processing

In mathematics, we can solve abstract problems by drawing it. This indicates that human beingsare better at understanding things by eyes. Heart Sound signal is one of the most important humanphysiological signals. We can convert abstract sound signal into a waveform image that people aremore adept at understanding. The two-dimensional phonocardiogram concept of this paper is basedon this theory; it is this paper’s unique innovation.2D-PCG can truly record of normal Heart Sounds,extra Heart Sounds, and cardiac murmur. In medicine,2D-PCG and cardiac auscultation can learnfrom each other. In the physiological information,2D-PCG contains rich personal physiologicalinformation, such as personal health information and identity information.Image processing technology has matured after years of rapid development. Firstly, usingmethods of one-dimensional signal processing to realize wavelet noise reduction and amplitudenormalized of Heart Sound, and converts it to2D-PCG with unity and comparability. Then werealize pretreatment of2D-PCG by image processing, including Gray Processing, BackgroundNormalization, Binarization, and Refinement. Then we analyze of characteristics of2D-PCG,combining the physiological significance of Heart Sounds and image features of2D-PCG. What wefocused on is the vertical and horizontal coordinate’s ratio and the inflection point sequence code,Of which the vertical and horizontal coordinate’s ratio is proposed based on Heart Sounds’amplitude and time ratio, the inflection point sequence code which can characterize detailedfeatures of the identity information, is raised based on Heart Sounds’ bimodal wavelet, trimodalwavelet, quadrumodal wavelet concept. The characteristics of the wavelet decomposition are thelow-frequency coefficient matrix of image overall and high-frequency coefficient matrix of imagedetails.At last, we discuss the feasibility of the classification and identification of2D-PCG by SVMand Euclidean Distance. According to the data from the experimental results, the three features canachieve the classification of2D-PCG, of which the inflection point sequence code has the highestrecognition rate. In identification, the inflection point sequence code also has the highestrecognition rate. The results of this study show that the classification and identification of2D-PCGhas some feasibility and safety.2D-PCG has a broad application prospects.