论文部分内容阅读
听诊是通过听取心脏所发出的声音来帮助诊断各种心脏疾病的一种有效手段。鉴于目前机械瓣的使用非常普遍,研究简单有效的机械瓣病变判别方法对于临床诊断来讲具有很大的意义。针对五种不同的机械瓣心音进行的分析表明,运用频谱仅能鉴别瓣周漏这一种机械瓣病变。虽然直接利用信号的时频成分进行机械瓣心音分类是可能的,但识别率只有84.0%。利用改进的局部最优基(LDB)算法来提取特征对机械瓣心音分类有着非常大的帮助,识别率达到了97.3%。与原始的LDB算法相比,实验表明改进后的LDB算法对提高识别率和降低计算复杂性都有着明显的优势。
Auscultation is an effective means of diagnosing various heart diseases by listening to the sound of the heart. In view of the fact that the use of mechanical valves is very common at present, it is of great significance to study the simple and effective method of discriminating mechanical valve diseases in clinical diagnosis. Analysis of five different mechanical heart sounds shows that using the spectrum can only identify a mechanical valve lesion of the valve flap leakage. Although it is possible to classify the mechanical heart sounds directly using the time-frequency components of the signal, the recognition rate is only 84.0%. Extracting the features by using the improved local optimal basis (LDB) algorithm is very helpful for classification of mechanical heart sounds, with a recognition rate of 97.3%. Compared with the original LDB algorithm, experiments show that the improved LDB algorithm has obvious advantages for improving the recognition rate and reducing the computational complexity.