The Multiscale Sequential Filter with Multisensor Data Fusion

Abstract

Combining the multiscale capability from wavelet with the performance of real-time and recursion about Kalman filter, a multiscale sequential filter is proposed to process dynamic systems with multisensor. This filter can not only absolutely achieve the effect obtained via conventional multisensor fusion approach, but also it has the advantages as wavelet and Kalman filter. Its multiscale characteristic can be used to analyze stochastic signal in different frequency subspace. Some similar methods existed do not possess these capabilities, such as real time and recursion. Computer simulation also shows that all estimate results from the new algorithm is comparable with that from traditional date fusion algorithms. Finally, the computable advantage is likewise validated by comparing the computer burden between the new algorithm and other two existed fusion algorithms