Ventricular fibrillation is considered the most common cause of sudden cardiac arrest. The fibrillation, and ventricular tachycardia often preceding it, are cardiac rhythms that may respond to emergency electroshock therapy and return to normal sinus rhythm when diagnosed early after cardiac arrest with the restoration of adequate cardiac pumping function. However, manually checking ECG signals on the existence of a pattern of such arrhythmias is a risky and time-consuming task in stressful situations and practically impossible in the absence of a qualified medical specialist. Therefore, systems of the computer classification of arrhythmias with the function of making a decision on the necessity of electric cardioversion with the parameters of a high-voltage pulse calculated adaptively for each patient are widely used for the automatic diagnosis of such conditions. This paper discusses methods of analyzing the electrocardiographic signal taken from the electrodes of an external automatic or semi-automatic defibrillator in order to make a decision on the necessity for defibrillation, which are applicable in the embedded software of automatic and semiautomatic external defibrillators. The paper includes an overview of applicable filtering techniques as well as subsequent algorithms for extracting, classifying and compressing features for the ECG signal.