Abstract： Given the poor source reconstruction results of minimum variance beamforming(MV)for magnetocardiogram(MCG)at low signal-to-noise ratio in coronary heart disease patients,a novel method using the improved minimum variance beamforming(IMV)is proposed.Based on MV,the proposed method uses the lead matrix and the magnetic field signal second-order characteristic matrix to form a constraint matrix,and constructs the weight matrix of a new spatial filter,thus reducing the output noise power gain.The single-source resolution of IMV and MV spatial spectrum estimations is compared using the point spread function.IMV and MV are used to perform source reconstruction from the 36-channel magnetic field simulation data generated by 3 current dipoles at low signal-to-noise ratio;and the source reconstruction is also carried out on the R-wave peak and T-wave peak time data of 36-channel MCG of 3 coronary heart disease patients.The results show that IMV has higher single-source resolution and reconstruction accuracy for simulation data and MCG data from coronary heart disease patients.