We present acomplementary least-mean-square algorithm of adaptive filtering for SQUID-based magnetocardiography, in which both rapid convergence and fine tracking are realized by switching the weight parameters back and forth between two filters according to the least mean square principle.