基于EKF+EKS的BCG动态高斯模型滤波研究

BCG DYNAMIC GAUSS MODEL FILTERING BASED ON EKF+EKS

心冲击图(Ballistocardiogram,BCG)信号属于微弱信号,现实采集的BCG信号通常包含采集环境干扰和个体差异,因此缺乏健康个体的BCG信号模板和适用于BCG信号的降噪方法。提出一种基于高斯核函数的动态BCG信号模型,在动态高斯模型的基础上,应用扩展卡尔曼滤波与扩展卡尔曼滤波平滑对BCG信号进行复合滤波。该模型提供健康个体的BCG信号,包括H波、I波、J波、k波、L波和M波特征。经联合扩展卡尔曼滤波与扩展卡尔曼平滑滤波后的BCG信号比其他传统滤波器滤波后的BCG信号信噪比更高。基于动态高斯模型合成的BCG信号能完整表达健康个体的BCG信号特征,联合扩展卡尔曼滤波和扩展卡尔曼平滑的复合滤波,对BCG信号的滤波达到了更好的降噪效果。

BCG signal belongs to weak signal,and the collected BCG signal usually contains interference from the acquisition environment and individual differences.So there is a lack of healthy individuals BCG signal template and the methods suitable for BCG denoising.A dynamic BCG signal model based on Gauss kernel function was proposed.Based on the dynamic Gauss model,we proposed the extended Kalman filtering and the extended Kalman smoothing to compound the BCG signal.The model provided BCG signals of healthy individuals,including H wave,I wave,J wave,K wave,L wave and M wave characteristics.The signal-to-noise ratio of BCG signal filtered by joint extended Kalman filter and extended Kalman smoothing filter was higher than that filtered by other traditional filters.The BCG signal based on the dynamic Gauss model can fully express the BCG signal characteristics of the healthy individual.The combined extended Kalman filter and extended Kalman smoothing filter can achieve better denoising effect for BCG signal filtering.