Magnetocardiography (MCG) is an effective modality for clinical application and health monitoring due to non-contact measurement and mapping of heart activity at high spatial resolution. A superconducting quantum interference device (SQUID) magnetometer is usually used for measuring MCG signals. However, a SQUID magnetometer has high running cost due to the liquid helium. Moreover, measuring MCG signals inside a magnetically shielded room (MSR) can be costly. Therefore, we developed a 64-channel magneto-impedance (MI) sensor system that does not require an MSR. However, the MCG measurement has very high noise level without an MSR. In this paper, we discuss the signal processing techniques of various noise reduction methods to decrease very loud noises. In particular, we investigated three spatial lter conditions that decrease correlated noises among the 64-channel signals to achieve a high peak value of MCG signals. By using a spatial lter that uses the average of the circumference channels and gradient, the distortion of MCG signals can be reduced. The average reduction in amplitude of the R wave as a result of using a spatial lter was 4.5 pT. Furthermore, the signal to noise ratio (SNR) of the P wave was 29.1 dB, while that of the R wave was 42.3 dB, and clear MCG signals were obtained when using the spatial lter that uses the average of the circumference channels and the gradient. Finally, we successfully measured the MCG signals without an MSR.