System for reconstruction and visualization of cardiac electric activity, with high-resolution capabilities, which improves the performance of current systems by means of the inclusion of the a priori information of the bioelectric problem in a learning algorithm, and by means of the exploitation of the spatial and temporal correlations thanks to interpolation and decimation subsystems. The system consists of, at least: a plurality of catheters; ways for obtaining the location coordinates of said array; ways for obtaining the cardiac walls geometry; ways for processing the signals from the catheters, implementing an algorithm based on Support Vector Machines, allowing the incorporation of the bioelectric phenomena; ways of interpolation and redundance exploiting for yielding high resolution without matrix inversions; and ways of visualization.