基于功率谱信息熵的异常心音识别

目的：异常心音识别是心血管疾病检测的一种重要手段，为了探究异常心音频域的有用信息，提出了将不同频段的功率谱作为一个独立信源计算其信息熵，从而对房室瓣和动脉瓣异常信号进行判别的一种新方法。方法：实验先将心音信号进行小波包分解，然后利用改进的Welch方法计算信号的功率谱，进而求各频段的功率谱信息熵，再建立支持向量机预测模型来对两种异常心音进行识别。结果：选取二尖瓣狭窄、二尖瓣关闭不全、主动脉瓣狭窄、主动脉瓣关闭不全共27例心音信号进行算法仿真，其中房室瓣异常能够全部检测出来，动脉异常有3例被误判，正确率达到77％；在原有27例信号的基础上，增加3例房室瓣异常和3例动脉异常信号进行算法验证，房室瓣异常信号仍然能够全部被检测出来．动脉异常信号2例被误判。结论：从仿真结果可以看出，该算法对房室瓣异常和动脉异常两种心音信号有较高的识别率。尤其对房室瓣杂音能够完全识别，也表明功率谱信息熵在异常心音的识别中具有重要意义。

Recognition of Abnormal Heart Sounds Based on Power Spectrum Information Entropy

Objective： Recognition of abnormal heart sounds is one of the best important methods to detect cardiovascular system disease. A new algorithm, that could be used to recognize the abnormal signals from atrial-ventricular valve and aortic valve through computing power spectrum information entropy of every sub-frequency band, was presented, in order to explore the useful information in the frequency domain in abnormal heart sounds.Methods： The steps were as follows： First a wavelet pack- age was used to disassemble the original signals, then the power spectrum was calculated with the improved Welch, and then the power spectrum information entropy was calculated, at last a prediction model of SVM was proposed to distinguish the two sig- nal types. Results： 27 signals were used for algorithm simulation, including four kinds of abnormal signals such as mitral steno- sis, rnitral regurgitation, aortic stenosis and aortic regurgitation. All of the signals were judged right accurately except three ab- normal aortic valve signals. The detection average accuracy is 77%. Further in this experiment, another 6 signals were choosed to verify the algorithm. It turned out that all of the abnormal atriventricular valve signals could still be detected, but just one aor- tic valve signal determine right.Conclusions： The results show that the two types of signals could be well identified, especially the signals from atrial-ventricular valve. It can be seen that power spectrum information entropy plays an important role in the recognition of abnormal heart sounds.