CMStatistics 2022: Start Registration
View Submission - CMStatistics
B0920
Title: Semi-parametric independent component analysis in the time-frequency domain Authors:  Seonjoo Lee - Columbia University/New York State Psychiatric Institute (United States) [presenting]
Abstract: Independent component analysis (ICA) is a blind source separation method to recover source signals of interest from their mixtures. Most existing ICA procedures are for independent sampling assumptions and are carried out by estimating the marginal density functions. Second-order statistics-based source separation methods have been developed based on parametric time series models for the mixtures from autocorrelated sources. However, when the sources have temporal autocorrelations with mixed spectra, the second-order statistics-based methods cannot separate the sources. To address this issue, we propose a new ICA method by estimating spectral density functions and line spectra of the source signals using cubic splines and indicator functions, respectively. The mixed spectra and the mixing matrix are estimated via maximizing the Whittle likelihood function. Then, we extended this method for the non-stationary sources using time-frequency domain analysis. We illustrate the performance of the proposed method through simulation experiments and a resting stage EEG data application. The numerical results indicate that our approach outperforms existing ICA methods, including SOBI algorithms.