Robust Singular Spectrum Analysis
Wenxi Zhu
Abstract
As an adaptive signal decomposition method, singular spectrum analysis has been affected by sparse outliers, which bring serious deviation. In order to correct the singular spectrum distribution, we study a novel robust singular spectrum analysis method. Convex relaxation robust principal component analysis and Hankel matrix structured constraints construct robust singular spectrum analysis (RSSA) to recover potential singular spectrum and estimate sparse noise. RSSA can be used as a preprocessing method, and improve the robustness of subsequent applications to outliers. Preliminary experiments show the effectiveness of RSSA.
Topics & Concepts
Singular spectrum analysisRobust principal component analysisRobustness (evolution)OutlierSingular valueSingular value decompositionPrincipal component analysisComputer scienceMathematicsAlgorithmArtificial intelligenceEigenvalues and eigenvectorsPhysicsQuantum mechanicsBiochemistryGeneChemistryStatistical and numerical algorithmsImage and Signal Denoising MethodsStructural Health Monitoring Techniques