Blind Separation of Nonstationary Sources by Spectral Decorrelation

Abstract. This paper demonstrates and exploits some interesting frequency-domain properties of nonstationary signals. Considering these properties, two new methods for blind separation of linear instantaneous mixtures of mutually uncorrelated, nonstationary sources are proposed. These methods are based on spectral decorrelation of the sources. The second method is particularly important because it allows the existing time-domain algorithms developed for stationary, temporally correlated sources to be applied to nonstationary, temporally uncorrelated sources just by mapping the mixtures in the frequency domain. Moreover, it sets no constraint on the variance profile, unlike previously reported methods.

• Type: Conference
• Year: 2004
• Where: ICA
• Authors: Shahram Hosseini, Yannick Deville
• Keywords: ICA 2004, Interesting Frequency-domain Properties, Linear Instantaneous Mixtures, Uncorrelated Sources