| Congenital heart disease (CHD) belongs to birth defect. Which Can be cured by operation. furthermore, early detection and treatment can improve the recovery rate. While the early symptoms of CHD is not obvious,and the clinical diagnosis of CHD is being divided into initial diagnose and confirmed diagnose, the initial diagnose are mainly relied on the auscultation, which is very related to the subjective experience from doctor, and delay the previous time to get a better treatment. Phonocardiogram is a digital picture and have much information about the activity of heart. So the disadvantages of diagnose making would be improved with the digital signal processing to Phonocardiogram, and rise the diagnosis efficiency.Starting from the generation mechanism of heart sound signal frequency characteristics, as well as the basic heart sound procedure, including pretreatment, heart sounds segmentation, feature extraction. This paper gives quantitative analyses of heart sound signals.In the pretreatment stage, This paper raises simplified multi-resolution decomposition of wavelet packet algorithm in view of the heart sound analysis. From this, we can see the energy distribution of heart sound in the range of the, different frequency, effectively separate the noise and heart sounds, physiological and pathological noises, then combined with the relationship between the normal heart sound and the heart noise, we dividing the coefficient of wavelet packet into four frequency bands qualitatively, it refers to lower, low, middle, high frequency, then we choose Shannon energy algorithm to extract the envelope, and according to the self-correlation principle, we highlight and compress the four envelope by multiplying and adding, and get heart sound envelope and scale envelope. The heart sound envelope reflects the overall envelope. The scale envelope owns the advantage of optimize the singular point of signals and weaken the lower energy rag in the envelope.In the heart sounds segmentation stage, this paper puts forward a new envelope extraction, using the heart sound envelope and the scale envelope, combined with the basic feature of heart sounds in times-domain, realizing the adaptive segmentation and position of heart sounds. Without ECG and other reference information, the method gets the envelope information completely and accurately from the time-domain and frequency-domain. the positioning and segmentation accuracy above95%for the50cases of normal and abnormal of complex heart sounds. New method is not only high efficiency, at the same time,it could reduce the possibility of wrong boundary positioning by moving a fixed length around the two sides of peak.In the terms of feature extraction, this paper has carried on two aspects of heart sounds feature extraction, including noise detection and heart reserve evaluation. we use the cardiac reserve index (S1/S2, D/S, HR) proposed by Professor Xing-ming Guo for non-invasive detection of congenital heart disease, and found the D/S rating from the abnormal sound is low, while the HR is high. In addition, the appearance of heart noise is the most evidence which heralds the heart disease. For this, this paper extracts the diastolic and systolic period energy fraction, certifies that the energy fraction of abnormal heart sound are higher than the normal ones in the middle-band and high-band, which foresee the pathologic murmur of heart sound. Take this all into consideration, according to the analysis and processing of the heart sound, the outcome shows the potential application of the further clinical disease diagnose in the future.
|
PDF Full Text Request