基于短时傅里叶变换的心音特征提取研究
Research on Heart Sound Feature Extraction Based on Short-Time Fourier Transform

针对正常与异常心音在时频域具有不同特征表现形式，本研究提出了一种基于短时傅里叶变换结合等高线轮廓的方法对心音信号进行时频域特征提取研究。首先，基于小波分解对心音信号进行预处理，保留21.5~689 Hz频段有效心音信号；其次，基于短时傅里叶变换对有效心音进行时频分析；最后，本研究分析了40位健康志愿者的1500秒正常心音信号并与典型异常心音信号进行比较，其结果表明：正常第一心音频带宽度为70.07 ± 12.62 Hz，持续时间为109.3 ± 26.2 ms。PS (肺动脉狭窄)病人第一心音频带宽度为168.80 ± 9.95 Hz，持续时间为23.7 ± 3.0 ms。
Aimed at the different characteristics of normal and abnormal heart sounds in the time-frequency domain, this study proposes a short-time Fourier transform (STFT)-based method combined with manual selection method to extract the features of heart sounds in the time-frequency domain. Firstly, the heart sound signal is preprocessed via wavelet decomposition, and the effective frequency components (21.5 - 689 Hz) of heart sound are remained. Secondly, STFT-based heart sound features are defined and extracted to character heart sound. Finally, the 1500 seconds normal heart sounds signal from 40 healthy volunteers are analyzed to compare with several typical abnormal heart sounds. The comparative results show that the frequency width and time width of normal sounds are distributed in 70.07 ± 12.62 Hz and 109.3 ± 26.2 ms, respectively. However, the time and frequency widths distributed 168.80 ± 9.95 Hz and 23.7 ± 3.0 ms are corresponding to the sounds from patients with pulmonary stenosis.