In order to analyze the feature of heart sound accurately and effectively, this paper presents a feature analysis approach of heart sound based on the improved Hilbert-Huang Transform after a large number of analysis of heart sounds in time frequency domain. The validity of the proposed method has been verified through Empirical Mode Decomposition (EMD) for a typical vibratory. Calculating and obtaining the characteristic parameter of heart sound by Hilbert spectrum analysis of several cases of normal and abnormal heart sounds. Experimental results show that the presented algorithm is able to identify different heart sounds in time frequency domain, and it also establishes the basis for the classification and recognition of heart sound.