<p>To effectively identify and differentiate between different heart problems based on heart sound signals, and to assist clinicians in diagnosing various heart conditions. Heart sound signal analysis is crucial for the early diagnosis of heart diseases. A heart sound segmentation method is proposed that incorporates heart sound envelopes and time-frequency characteristics aimed at improving the precision of heart sound segmentation. Considering the effects of noise on the precision heart sound segmentation and feature envelopes on heart sound component extraction. The wavelet packet transform is used to decompose and reconstruct the heart sound signal, divide the effective signal from the noise signal to the greatest extent possible, and reduce noise signal interference. Meanwhile, hilbert envelope noise interference is severe, the signal boundary burr is heavier, and it is difficult to accurately extract the signal components. This paper introduces the first-order Shannon energy to enhance the envelope signal, reduce the burr, and highlight the signal components. The PhysioNet/CinC Challenge results show that the method performs 15.7% better for normal heart sound signals and 18.7% better for abnormal heart sound signals than the baseline method. The robustness of the method to different Gaussian white noise contaminations is also demonstrated.</p>