Ventricular fibrillation (VF) and ventricular tachycardia (VT) are the foremost reason of sudden cardiac deaths which can be cured by the timely application of automated external defibrillators (AEDs) by identifying the sternness of cardiac arrhythmias using complex algorithms. Since heart disease is a chronical disease which cannot be cured permanently, hence an early prediction of these diseases is very much crucial. In this work, we have suggested a novel algorithm using DWT based VMD features. Primarily, DWT has been applied on the ECG signal using Daubechies db6 mother wavelet to decompose it up to 8th level. Then on each level, the VMD algorithm has been applied to extract a set of 15 time-frequency features. The computed features are then used for identification and precise classification of VT, VF, and normal rhythm. This feature set was authenticated with the estimated data available in the CUDB and VFDB database obtained from PhysioNet database. The proposed algorithm results in an accuracy of 99.13% in CVR classifier and 93.88% in SVM classifier. The results show that the t-f data representations that are fed to the classifiers provide superior performance values.