Extension of Power Method to Para-Hermitian Matrices: Polynomial Power Method
- Resource Type
- Conference
- Authors
- Khattak, Faizan A.; Proudler, Ian K.; Weiss, Stephan
- Source
- 2023 31st European Signal Processing Conference (EUSIPCO) European Signal Processing Conference (EUSIPCO), 2023 31st. :1564-1568 Sep, 2023
- Subject
- Signal Processing and Analysis
Simulation
Signal processing algorithms
Europe
Estimation
Eigenvalues and eigenfunctions
Broadband communication
Sensors
- Language
- ISSN
- 2076-1465
This document extends the idea of the power method to polynomial para-Hermitian matrices for the extraction of the principal analytic eigenpair. The proposed extension repeatedly multiplies a polynomial vector with a para-Hermitian matrix followed by an appropriate normalization in each iteration. To limit the order growth of the product vector, truncation is performed post-normalization in each iteration. The method is validated through simulation results over an ensemble of randomised para-Hermitian matrices and is shown to perform significantly better than state-of-the-art algorithms.