基于S变换的心音信号特征提取

Heart sound feature extraction based on S transformation

为了准确获得心音信号生理、病理信息,在对比研究各种时频分析方法的基础上,采用具有良好时频分辨率的S变换提取心音信号特征,实现心脏病辅助诊断。首先对心音信号进行包络提取、第一心音定位及心音分段,然后对一个周期信号进行S变换,最后提取变换后矩阵的统计学特征,并将该特征作为区别不同心音信号的特征向量。通过相关评价可知该方法可以有效识别不同心音信号。

A new time-frequency analysis method for human heart sound was presented based on comparative study on many kinds of time-frequency analysis methods. Adopting S-transformation, it has good time-frequency resolution to display features of different heart sounds in order to assist heart diseases diagnosis. Firstly, with this method the heart sound envelope was extracted, the first heart sound was located and heart sound was divided into segments. Then, a periodic signal was transformed with S-transformation. Finally, the statistic features of the obtained matrix after S- transformation were extracted, these statistic features were taken as engenvectors to distinguish different heart sound signals. The evaluation results showed that this method can recognize different heart sounds effectively.