The emission regions of accreting X-ray pulsars provide a natural laboratory for physics under extreme conditions. However, various factors have prevented this potential from being fully exploited so far. Two of these factors are 1) the angular dependence of the cross-sections defining the radiative transport in the presence of strong magnetic fields and 2) the fact that we can observe emission from both poles of the neutron star simultaneously at some or even all phases of the pulse cycle. Coupled together, the observed emission is the result of two independent and intrinsically complex emission regions that is notoriously difficult to interpret. We have developed a novel technique to address this problem and decompose contributions of individual poles based on the observed pulse-to-pulse variability properties of pulsar light curves. Here we present the technique that is based on the Blind Source Separation problem, first results of its application to RXTE observations of Cen X-3, and discuss the comparison with earlier work aimed at solving the same problem.