베이지안 준지도 학습을 이용한 웨이블릿 기반의 시그널 노이즈 제거에 관한 연구

Abstract

Bayesian Adaptive Multi-resolution Shrinkage(BAMS)는 베이지안 법칙 기반의 효과적인 시그널 노이즈 제거 방법론의 하나이지만 웨이블릿 계수들의 구조에서 발생되는 정보를 활용하지 못하는 단점이 있다. 본 연구에서는 이러한 BAMS의 단점을 데이터 구조에서 나오는 정보를 효과적으로 활용하여 예측 성능 향상에 도움을 주는 준지도 학습을 이용해 보완하여 전체적인 시그널 노이즈 제거 능력의 향상을 목적으로 하는 새로운 모델을 제안한다. 연구의 진행을 위하여 관련 된 매개변수를 조정하여 사전에 정의되는 두 개의 BAMS룰을 만든다. 그리고 각 웨이블릿 계수들은 각각의 크기뿐 아니라 웨이블릿 계수들이 가지는 구조적인 특성을 고려하여 준지도 학습에 의하여 둘 중에 어떤 룰을 따를지 결정하게 된다. 본 논문에서는 제안 된 모델의 이론적인 특성에 대해 논의하고 적합한 매개변수의 값을 검증하여 제시한다. 또한 실험을 통하여 기존의 시그널 노이즈 제거 방법론들에 비해 제안된 모델이 더욱 뛰어난 성능을 보이는 것을 확인한다.|We propose an enhanced wavelet-based denoising methodology based on the Bayesian adaptive multiresolution shrinkage which is an effective Bayesian shrinkage rule plus the semi-supervised learning mechanism. The Bayesian shrinkage rule is advanced by utilizing the semi-supervised learning method in which the neighboring structure of a wavelet coefficient is adopted and an appropriate decision function is derived. According to the decision by the decision function, wavelet coefficients follow one of two prespecified Bayesian rules obtained by varying related parameters. The decision of a wavelet coefficient depends not only on its magnitude but also on a neighboring structure on which the coefficient is located. We discuss the theoretical properties of the suggested method and provide recommended parameter settings. We show that the proposed method is often superior to several existing wavelet denoising methods through extensive experimentation.; We propose an enhanced wavelet-based denoising methodology based on the Bayesian adaptive multiresolution shrinkage which is an effective Bayesian shrinkage rule plus the semi-supervised learning mechanism. The Bayesian shrinkage rule is advanced by utilizing the semi-supervised learning method in which the neighboring structure of a wavelet coefficient is adopted and an appropriate decision function is derived. According to the decision by the decision function, wavelet coefficients follow one of two prespecified Bayesian rules obtained by varying related parameters. The decision of a wavelet coefficient depends not only on its magnitude but also on a neighboring structure on which the coefficient is located. We discuss the theoretical properties of the suggested method and provide recommended parameter settings. We show that the proposed method is often superior to several existing wavelet denoising methods through extensive experimentation.