An adaptive best fitting plane algorithm for Stokes
space based polarization demultiplexing in digital coherent optical
receivers is presented. The algorithm allows to use the 3D
Stokes space representation of light polarization for polarization
demultiplexing in systems whose channel properties are time
varying. This is archived by updating the orientation of the
normal to the plane, sample by sample, which is used to
compute the channel inverse matrix. The effectiveness of the
proposed adaptive best fitting plane algorithm preserves the
high converge ratio and transparency to higher-level modulation
format properties of the original Stokes space based polarization
demultiplexing technique.