Computing electrical source in the heart from potentials on the body surface is an ill-posed problem of electrocardiogram (ECG). It is so difficult to get an exact solution of this problem. To solve this ECG inverse problem, a numerical analysis is used for forward problem
with a specific source model. Then, the solutions of ECG inverse problem is reached by using an iterative technique, Newton or downhill simplex, to find the best source whose potentials best match to the observed potentials in a least square sense. Such methods, however, often converge to
a local minimum and their results are affected by initial parameters. In this paper, source is considered as a moving single equivalent dipole and genetic algorithm, an efficient and robust
optimization method, is proposed to estimate parameters of source. In addition, some improvements are introduced to enhance performance of conventional genetic algorithm. A 3D volume conductor model of human body is constructed base on an anatomic atlas for numerical test. A comparison between our approach and one using downhill simplex method is implemented. The results show that our approach is stable and may provide a good scheme for solving the ECG inverse problem.
