<;p>;A method and system for detecting and classifying cardiac arrhythmias from the ECG signal source is disclosed. The method classifies arrhythmia in two stages, a coarse classification stage and a fine classification stage. The one or more input ECG beats obtained from the processed ECG signal. The input ECG beat is decimated and reference ECG beats from plurality of trained databases are decimated and compressed before classifying in coarse classification stage. The plurality of databases corresponds to a different number of arrhythmia classes. The coarse classification stage identifies one or more surviving classes upon calculating the distance between input ECG beat and reference ECG beats and the identified one or more surviving classes are forwarded to the fine classification stage for further accurate classification and thereby identifying one or more output classes corresponding to input ECG beat. FIG. 1<;/p>;