In order to solve the problem of power line noise and environmental magnetic noise to magnetocardiogram, a filtering method based on singular value decomposition (SVD) and generalized S transform is proposed. On the basis of discrete condition, S transform and its inverse transform in matrix form are derived. The signal is first preprocessed using generalized S transform in order to get better time-frequency resolution, then the time-frequency domain is decomposed by SVD, and the valid time-frequency region is determined. Therefore, the adaptive filtering can be realized. Numerical simulation studies indicate that the method presented can achieve better performance than other methods with less singular values. Illustrative examples verify the effectiveness of the proposed method.