How to extract the magnetocardiography (MCG) signals from the background noises and attenuate the noise is the key element of MCG　signal processing. From the aspect of software technique, a new adaptive algorithm is proposed to suppress the environmental magnetic noise of MCG in this paper. The propo sed algorithm is called improved complementary pair normalized least mean square algorithm. This adaptive structure is designed for MCG. This new algorithm prov ides the faster convergence speed and smaller steady-state error. In order to test the performance of this algorithm, the result of the simulated experiment and real MCG signal is analyzed. This method bring about the effective noise atten uation. Meanwhile, this algorithm is suitable for noise cancellation of other non-stationary signals.