心脏疾病是全球发病率和死亡率最高的疾病,心音听诊可以获取心脏的机械特性及结构特征,与超声心动图、核磁共振等无创诊断技术相比具有快速、低成本和操作简单的优势.心音信号成分复杂,容易受到各种噪声和干扰的影响,听诊诊断结果容易受到医生主观性的影响,极大限制了心音听诊的应用.该文提出一种基于心动周期估计的心音分割及异常心音筛查算法,预先估计了心音的心动周期,存在随机干扰的情况下也可以正确识别信号中80%以上的心动周期,提高了算法的稳定性.同时提出了区分度良好的时域和频域特征指标,利用支持向量机建模,对异常心音的识别率可达92%.算法可辅助医生诊断,或用于家用便携式心音监护设备.%Heart disease is of highest morbidity and mortality. The cardiac structure and mechanical characteristics can be reflected by auscultation. Compared with echocardiography and nuclear magnetic resonance, auscultation gets the advantages of fast, low cost and easy to use. The composition of phonocardiogram is complex, and the auscultation is easy to be affected by the subjectivity of the doctor, various noise and disturbances, which limits the application of auscultation. The algorithm of phonocardiogram segmentation and abnormal phonocardiogram screening is presented. For the reason that the heart cycle is estimated in advance, 80% cardiac cycle can be recognition correctly when random disturbances exist. The diagnostic indexes of time and frequency domain with high discrimination are also presented, and the abnormal heart sounds are recognized by Support Vector Machine (SVM) with the accuracy about 92%. The algorithm can be used for assisting doctors or portable phonocardiogram monitoring device.
展开▼
