A new model for describing electrocardiography (ECG) is presented, which is based on multiple dipoles compared to standard single dipole approaches in vector electrocardiography. The multiple dipole parameters are derived from real data (e.g. four dipoles from 12-channel ECG) by solving the backward problem of ECG numerically. Results are transformed to a waveform description based on Gaussian mixture for every dimension of each dipole. These compact parameterized descriptors are used for a very realistic real-time simulation applying the forward solution of the proposed model.