Bayesian information criterion for multidimensional sinusoidal order selection
Jie Xiong, Kefei Liu, João Paulo Carvalho Lustosa da Costa, Wen-Qin Wang
Abstract
Detecting the sinusoidal order is a prerequisite step for parametric multidimensional sinusoidal frequency estimation methods, whose applications range from radar and wireless communications to nuclear magnetic resonance spectroscopy. Although the Bayesian information criterion (BIC) has been commonly applied for model order selection, its application to sinusoidal order estimation is recent. By means of estimation of Fisher information matrix, we extend the 1-D BIC to multidimensional case for multidimensional sinusoidal order selection. The multidimensional BIC is shown in simulations to outperform the state-of-the-art algorithms in terms of probability of correct detection.
BibTeX
@inproceedings{icassp2017_bayesianinformat,
title = {Bayesian information criterion for multidimensional sinusoidal order selection},
author = {Jie Xiong and Kefei Liu and João Paulo Carvalho Lustosa da Costa and Wen-Qin Wang},
booktitle = {ICASSP 2017},
year = {2017}
}