基于改进NMF与相位补偿的胎心音降噪算法

FETAL HEART SOUND DENOISING BASED ON IMPROVED NMF AND PHASE SPECTRUM COMPENSATION

在孕妇体表,通过声音传感器对胎儿心音信号进行采集,结果通常包含多种噪声,常用的滤波方法无法取得满意的降噪效果。为了能够更好地滤除与胎儿心音信号同频的噪声信号,提出改进的非负矩阵分解算法(NMF),对短时幅度谱利用奇异值分解得到用于NMF的特征数和初始化矩阵,对NMF增加L2,1稀疏限制,使分解得到的胎心音特征呈现更多细节;对短时相位谱采用了增加稀疏限制的相位补偿算法(PSC)。实验结果表明:与常用降噪方法相比,该方法的信噪比至少提升0.52 dB,能更好保留胎儿心音信号特征。

The fetal heart sound signal is collected by the sound sensor on the body surface of pregnant women,and the result usually contains a lot of noise.The commonly used filtering methods cannot achieve satisfactory noise reduction results.In order to better filter out the noise with the same frequency as the fetal heart sound signal,an improved non-negative matrix factorization was proposed.For the amplitude information matrix,the singular value decomposition was used to calculate the feature number and the initialization matrix for NMF.In addition,the L2,1 sparse limit was added to the NMF to make the features of the fetal heart sound obtained by decomposition more detailed.The phase spectrum compensation(PSC)with sparse limit was applied to the phase angle information matrix.The experimental results show that compared with the commonly-used noise reduction methods,the signal-to-noise ratio of this method is improved by at least 0.52 dB,which can better preserve the characteristics of fetal heart sound signals.