基于高阶香农熵的心音分段算法

Heart sound segmentation algorithm based on high-order Shannon entropy

针对心音分段中存在的分段结果容易受心杂音干扰的问题,改进了基于香农能量的心音分段算法。先用小波变换对心音信号进行预处理,消除环境噪声和高频杂音的影响,再计算信号的高阶香农熵,并以此作为信号包络,抑制低频杂音的干扰,最后根据生理知识对包络进行分段,确定分段边界。利用本算法对包含正常和异常心音的实验数据进行分段,正确分段率达96%以上。

The algorithm based on Shannon energy is improved for robust segmentation of heart sound with murmurs. First, the heart sound signal is preprocessed by wavelet to eliminate background noises and high-frequency murmurs. Then, the high-order Shannon entropy of signal is calculated as envelope to overcome the interference of low-frequency murmurs. Finally, the envelope is segmented into four parts： the first heart sound, the systolic period, the second heart sound, and the diastolic period according to physiology knowledge; and the accurate boundaries of the segmentation are detected. The algorithm was tested using normal and abnormal clinical data. The results show that the correct ratio of the algorithm is over 96%.