A novel pre-processing technique of blind source separation applying Q-mode factor analysis

In this study, a novel way of processing observations before independent component analysis (ICA) using a Q-mode factor analysis (FA) was proposed for noisy blind source separation (BSS). The Q-mode analysis is a very efficient technique in classifying a data in cases where there are a large number of objects and where there is a little prior knowledge of the constituents. In the R-mode analyses, interrelationships between variables are analyzed. On the other hand, in the Q-mode analysis, interrelationships between objects are analyzed. Applying this approach to the experimental noisy data, we show that our proposed approach is more effective than the R-mode analysis for source separation of noisy data.