基于功率谱第三心音特征提取研究

PSD-Based Feature Extraction for the Third Heart Sound Analysis

摘要: 针对第三心音在频域上的复杂特征，本研究采用基于功率谱密度的频率特征提取方法，对第三心音的特征进行评价，并将正常和异常的第三心音进行区分。与本研究相对应的阶段安排如下：1) 首先通过电子听诊器采集心音，并基于小波分解对其进行预处理，消除背景噪声，保留有效信号。2) 人为选择的第三心音生成基于PSD的功率谱。然后，基于功率谱包络提取给定阈值线上的频率特征。3) 通过对正常第三心音和来自二尖瓣回流心脏病患者的异常第三心音对比分析，以阈值–(平均值 ± 标准差) Hz表示的统计结果显示，正常第三心音的特征分布在0.2~(56.0832 ± 6.3373) Hz、0.8~(21.0824 ± 2.2393) Hz，异常S3为0.2~(126.7094 ± 27.0634) Hz，0.8~(64.7820 ± 11.6584) Hz。

Abstract: In view of the complex characteristics of the third heart sound (S3) in the frequency-domain, this study employs a power spectrum density (PSD)-based frequency features extraction meth-od to evaluate the characteristics of the S3 and to discriminate normal from abnormal S3. The stages corresponding to this study are arranged as follows: 1) The heart sound is firstly collect-ed via an electronic stethoscope and is preprocessed based on wavelet decomposition to elimi-nate the background noise and retain the effective signal. 2) The PSD-based power spectrum is generated for the artificially selected S3. And then, an envelope for the power spectrum is based to extract the frequency features over a given threshold value (Thv) line. 3) By comparatively analyzing the normal S3 and abnormal S3 from the patient with mitral regurgitation heart dis-ease, the statistical results expressed as Thv-Frequency range (mean ± standard deviation) Hz show that the features of normal S3 are distributed in 0.2 - (56.0832 ± 6.3373) Hz, 0.8 - (21.0824 ± 2.2393) Hz, while those of abnormal S3 are in 0.2 - (126.7094 ± 27.0634) Hz, 0.8 - (64.7820 ± 11.6584) Hz.