基于活动性检测动态估计噪声的心音降噪算法

Heart sound denoising algorithm based on the dynamic estimation of noise via activity detection

针对基于小波分解和最优改进对数幅度谱估计的心音降噪算法存在噪声残留和心音失真的问题,提出一种基于心音活动性检测(HSAD)动态估计噪声的心音降噪算法。通过设计的HSAD判断当前心音帧是否为基础心音帧(FHS),根据判断结果分别采用改进最小值控制递归平均(IMCRA)算法和递归平滑算法对噪声功率进行动态估计与更新,采用非因果先验信噪比,实现心音信号的降噪。实验结果表明,提出算法能更好在提升降噪性能的同时,降低FHS的失真。

Aiming at the problems of noise residue and heart sound distortion in the wavelet decomposition based denoising algorithm and the optimally modified log-spectral amplitude estimator based denoising algorithm,an improved heart sound denoising algorithm based on dynamic noise estimation by the heart sound activity detection(HSAD)was proposed.The designed HSAD was used to judge whether the current heart sound frame was the fundamental heart sound(FHS)frame.According to the judgment results,the improved minimum control recursive average(IMCRA)algorithm and the recursive smoothing algorithm were used to dynamically estimate and update the noise power,at the same time,the noncausal prior signal-to-noise ratio was combined to complete the noise reduction of heart sound signal.Experimental results show that the proposed algorithm can improve the noise reduction performance and reduce the distortion of FHS.