This contribution presents the development of the pattern recognition procedure for ECG signal evaluation. In contrast to the methods already available, the procedure is not based on an ever deeper splitting of the signal into individual attributes but uses procedures of signal pattern recognition in conjunction with ECG databases. With the aid of an ECG signal pattern database with 12000 ECGs, sensitivities and specificities for the main groups of findings are presented and discussed for 8500 examinations. Conclusions for the further development of the procedure are drawn.