An adaptive version of the algebraic constant modulus algorithm [blind source separation applications]

The algebraic constant modulus algorithm (ACMA) is a very effective block-algorithm for blind source separation. A particular feature is that it finds beamformers for all sources simultaneously. However, in cases with continuously transmitting sources and varying environments, data-adaptive algorithms may be more appropriate. In this paper, we derive an adaptive version of the ACMA, with a reasonable computational complexity. Simulations indicate that it is reliably tracking beamformers to all sources even in a highly time-varying scenario.