1 Introduction
Although signal decomposition is a popular technique widely used in many branches of science and engineering, blind source separation (BSS) deals with a real-world situation where neither the sources nor the mixing matrix is known and the only available information comes from the mixed observations [1], [2]. For instance, multiprobe biomedical imaging exploits simultaneous imaging of multiple biomarkers, where the measured pixel values often represent a composite of multiple sources independent of spatial resolution (e.g., multispectral microscopy, dual-energy X-ray imaging, dynamic functional imaging, and electroencephalogram or magnetoencephalogram (EEG/MEG) [3], [4], [5], [6], [7], [8]). Other examples include remote sensing [1], astronomical imaging [9], analytical spectroscopy [10], and telecommunications [11]. A popular approach to BSS is the independent component analysis (ICA) [12], where the sources are fundamentally assumed to be mutually and statistically independent, although this fundamental assumption may hardly be true in many real-world problems.