基于多尺度Kalman数据融合滤波

本文通过分析基于小波变换的动态系统模型,提出一种基于小波多尺度的Kalman数据滤波方法,本文利用小波的多尺度特点,把初始估计序列多尺度分解,并在不同尺度层上进行Kalman滤波估计,再利用小波重构来融合各层的估计信息,把标准Kalman滤波只在单一尺度和时间轴上对状态估计值和误差协方差进行数据更新,改进为基于小波变换的尺度轴和时间轴上的双向数据更新,该算法将小波多尺度分解去噪和Kalman滤波相结合,对实际中含较强噪声的动态系统的状态估计效果较好.算法也可用于多分辨率多传感器数据融合.

Data Fusion Filtering Techniques Based on Multiscale Kalman Filter

In this paper, with analysis on the dynamic system model based on wavelet transform, a novel data fusion filtering algorithm based on multiscale Kalman filter was proposed. Wavelet transform was very suitable for processing multiscale signals, and it could be used in multiresolution fusion filtering. The approach used the multi-scale characteristic of Wavelet transform to decompose the primal forecasting sequence firstly, and performed Kalman filtering estimate on each level, and lastly wavelet reconstruction was applied to integrate the estimate information from neighborhood scale. So the better filtering precision is obtained. The wavelet decomposition of the noised measurements was combined with Kalman filter in every scale and it makes the best of multiresolution data information. Thus, the better filtering results are achieved for the dynamic system with high-level noise in practice.