次頻模式及其應用

Abstract

在這篇論文中，對於傳統ＡＲ模式我們提出了一個新的藉由次頻技術的方 法。經由這種方法可得到更有效率的實現及更好的性能。我們並將其應 用 ，濾波及形狀辨識。前兩種應用我們提出了在次頻模式 新方法，並且 為了實現簡單，又提出了零階趨近法。所 膆雃劃堣隤k的優越性。而形狀 辨識方面，用次頻ＡＲ為特徵也同樣地增進了辨識率。 In this thesis we propose an new method for the tranditional AR modeling problem via subband technique. By means of this method more efficient realization and better performance will then achieve. Three alppications including prediction, and shape recognition are explored. Unlike trainditional implementation of predicton and filtering, the algorithm we proposed is modeled in subband but realized in full band. For ease to implementation a zero-order method is proposed. All the simulation results indicate the preponderance of our algorithm. Pattern recognition that employ subband AR feature also enhance the correct ratio of classification.