Bioelectric signals provide important information about the physiological function of many living organisms. In magnetism, the so-called direct problem deals with the determination of the magnetic field associated to well known current sources. On the other hand, there are situations where it is necessary to determine the current source responsible for the generation of a measured magnetic field. This type of problem is common in Biomagnetism and is called inverse problem. For example, based on cardiac magnetic field measurements it is possible to infer the electrical activity in the heart tissue, responsible for its generation. This work proposes, presents and discusses a new technique designed to solve the biomagnetic inverse problem by genetic algorithms. It is intended to estimate the position, orientation and magnitude of the equivalent current dipoles, responsible for the generation of biomagnetic field maps measured with a 16 channel SQUID system. The algorithm attempts to identify the distribution of dipoles that best fits the measured experimental data, aiming at minimizing the error between the experimental magnetic field maps and those obtained by the estimated solutions. The experimental data analyzed in this study were acquired by measurements in isolated rabbit hearts. The knowledge of parameters of current dipoles at different instants of time allows the correct interpretation and analysis of medical information obtained from the experimentally measured biomagnetic fields, providing diagnosis and guiding therapeutic procedures.