Ventricular fibrillation (VF) is one of the primary causes of sudden cardiac arrest, a potentially lethal arrhythmia. The heart quivers rather than pumps normally due to this erratic cardiac electrical activity. Currently, the only proven treatments for VF are defibrillation and early cardiopulmonary resuscitation (CPR). VF needs immediate medical intervention. When VF is identified early and with accuracy, survival chances may increase significantly. With the use of Multi-Layer Perceptron (MLP) for binary classification and modified 1D U-Net architecture for feature extraction, this study proposes a novel model for VF identification. To precisely identify the VF in this case, Heart Rate Variability (HRV) traits are employed. The MLP employs hyperparameter adjustment and 5-cross validation (5-CV) to improve the outcomes. In terms of VF detection, the model has achieved 99.56% accuracy.