An ultra-highly sensitive magnetic sensor without refrigerant has a potential to realize a ubiquitous real-time measurement system of magneto-cardiogram (MCG) for automatic certified biometrics. We investigated (1) magneto-impedance (MI) sensor materials for increasing signals, (2) detection circuits for reduction of noises, and (3) certification methods from MCG signals by generative adversarial networks (GAN). Finally, pseudo MCG signals were detected by the MI sensing system and identified using GAN learned by ECG data set.