Deep learning (DL) models using Convolutional Neural Network (CNN) have proven to be very proficient in deriving complex patterns and characteristics from unprocessed electrocardiogram (ECG) data, allowing for the reliable classification of arrhythmias in different patient groups. Arrhythmias are a major problem in the field of cardiovascular health because of their wide range of symptoms and potentially fatal consequences. The primary focus of this study is to create a DL framework designed to precisely classify arrhythmias into five distinct ectopic categories found in ECG recordings and provide support to cardiologists. Cardiologists will find it easier to identify patients who may have a cardiac condition with this work since the updated characteristic replaces manually generated features and thereby assists in better diagnosis. The acquired data, from the arrhythmia dataset, is studied and proceeded for the required preprocessing which involves upsampling and downsampling. After understanding the preprocessed data by visualizing tools, to enhance model robustness, data augmentation is applied by adding Gaussian noise. The CNN architecture comprises three convolutional layers, batch normalization, pooling layers, and dense layers and its performance is evaluated using accuracy and visualizing its training history. Analyzing the model performance graph for both the training and validation data, the model effectively classifies arrhythmia categories with 93% accuracy, highlighting its potential as a valuable diagnostic tool for cardiologists. This approach effectively classifies ECG signals, identifying various ectopic beats.