Study On Heart Sound Denoising And Eigenvalue Extraction

The useful information related to heart disease can be reflected in the heart sound signals, such as coronary stenosis produce noise in heart sounds of diastolic. Therefore, heart sound detection for clinical doctors diagnosing heart disease has a supplementary role.In the course of recording heart sound, inevitably, there will be noises merging in the main signal. Before further processing the phonocardiographic records, noise must be suppressed first. Considering the nonstationarity of PCG, the application of time-frequency and wavelet transform in heart sound denoising were analyzed,and eventually adopted the method of double adaptive lifting wavelet transform, that the adaptive method to design update operator and predicting operator in the process of ascension. At the same time, an improved threshold function also was introduced in order to overcome the soft threshold function and hard threshold function faults.Noise must be suppressed for the phonocardiographic records with double adaptive ascension wavelet transform and improved threshold function,which proved to have a good performance.The selection of the characteristic value of heart sound signals is the key for properly automatic identification. In this paper, we extracted time domain and frequency domain of characteristic value for heart sound signal with HHT analysis, the normalized energy of the optimal basis with wavelet packet algorithm as the feature vector. The five types of heart sound signals are classified recognition according to the characteristic value with support vector machine,which recognition rate reached97.1%.