Heart sound signal can reflect the activity of the human heart valve, and the heart sound classification can distinguish the pathological information of different heart sounds, which is of great significance for clinical diagnosis of different heart diseases. Heart sound segmentation is the premise of heart sound classification. The heart sound segmentation can locate the first heart sound (S1) and the second heart sound (S2) in the heart sound, and provide a positioning reference for heart sound feature parameter extraction and heart sound classification. For this reason, a new adaptive threshold selection heart sound segmentation algorithm is proposed in this paper. This method first uses the wavelet transform default threshold method to denoise the heart sound signal; then uses the normalized Shannon energy to extract a smoother heart sound envelope; then performs effective peak detection on the envelope to determine the initial large threshold TH1, and the final stable double threshold is obtained by an iterative method; finally, heart sound segmentation and the segmentation result analysis are performed. For partial abnormal heart sounds segmentation results, such as the segmentation result of heart sound splitting, the heart sound segment is merged or removed by using the characteristics of heart sound time domain and energy, which ensures the accuracy of the segmentation result. The experimental results show that the segmentation accuracy of normal and abnormal heart sounds is 97.24% and 91.83%, and the overall segmentation accuracy is 95.56%, which is higher than the traditional threshold selection segmentation method.
