In view of the fact that BP algorithm based on the common used error square sum rule and Sigmoid transfer function have some limitation and shortcomings, the cross entropy rule and a new transfer function are adopted for constructing and training process of the fuzzy neural network classifier. It is used to realize the orientation of myocardial infarction, and the results prove that this classifier has the capability of outperforming the traditional fuzzy neural network in training efficiency and recognizing ability obviously.