Multiscale Autoregressive Stochastic Model And Its Implementation

The multiscale autoregressive (MAR)is newly introduced statistical framework, which incorporates the multiscale analysis idea and have solve many challenges in the field of statistical signal processing successfully.To harness the power of the MAR framework,of course,requires that the phenomena of interest be effectively modelled in the framework. Many person focus on the problem of model identification.This thesis develop another approach,which utilize fully the appreciable decorrelation feature of wavelet transform, has the minimal computation burden of all the known alternatives, and may be the key to solve the problem of realizing MAR from samples directly .Another focus of thesis is to discuss the model ability and its applications of our developed framework which has special structure .