Algorithms and structures for source separation based on the constant modulusproperty

We propose two structures and theirs associated algorithms designed to solve the blind source separation problem in the presence of noise and interferences. Both structures exploit the non convexity of the Constant Modulus cost function, finding its multiple local minima. A convergence analysis shows that both schemes achieve the desired solution, separately extracting the sources of interest while rejecting noise and interferences, provided that they do not share the constant modulus property.