Estimating Phase Linearity in the Frequency-Domain ICA Demixing Matrix

We consider a method for solving the permutation problem in blind source separation (BSS) by the frequency-domain independent component analysis (FD-ICA) by using phase linearity of FD-ICA demixing matrix. However, there is still remaining issue how we can estimate the phase linearity. In this paper, we propose two methods to estimate the linearity of the phase response of the FD-ICA demixing matrix. Our experimental result shows that our new methods can provide better estimation of the phase linearity than our previous method.

• Type: Conference
• Year: 2009
• Where: IDA
• Authors: Keisuke Toyama, Mark D. Plumbley
• Keywords: Artificial Intelligence, FD-ICA Demixing Matrix, Frequency-domain Independent Component, IDA 2009, Phase Linearity