针对心音特征具有随其可分割性而改变的性质，提出一种基于心音分割的自适应特征提取算法，进而创建一种创新型的心脏病诊断系统。其创新点主要体现在：基于短时修正希尔伯特变换的第一复杂心音(CS₁)，第二复杂心音(CS₂)或完整心音(CS)的自适应分割提取；基于分割心音的自适应频率特征FF1或FF2提取；基于主成分分析的多尺度特征[γ₁₁、γ₁₂]和[γ₂₁、γ₂₂、γ₂₃]降维处理。实现此研究目标的2个阶段概括为：① 自动统计分析两个连续峰值之间的时间间隔，以此来确定心音的可分割特性；② 基于心音分割的自适应特征提取以及降维处理。通过在线数据库和临床数据库中提取的心音特征的散点图对系统性能进行初步评估验证。
An adaptive feature extraction algorithm based on heart sound segmentation is proposed to create an innovative heart disease diagnosis system, which based on the nature of heart sound features that change with their segment. The innovations of this methodology are primarily reflected in the automatic segmentation and extraction of the first complex heart sound (CS₁), the second complex heart sound (CS₂) or each cardiac sound (CS) based on the short-time modified Hilbert transform; adaptive frequency feature FF1 or FF2 extraction based on the segmented heart sounds; and dimensionality reduction of the multiscale features [γ₁₁、γ₁₂] and [γ₂₁、γ₂₂、γ₂₃] based on principal component analysis. The two phases of this study are summarized as: ① automatic statistical analysis of the time interval between two consecutive peaks to determine the segmental characteristics of the heart sounds; ② adaptive feature extraction and dimensionality reduction based on heart sound segmentation. A preliminary evaluation of the performance of the system is validated by scatter plots of heart sound features extracted from the online and clinical databases.