Electrocardiogram (ECG) signals are vital physiological markers used in the diagnosis of cardiovascular diseases. It is crucial to have high-quality signals to ensure accurate and reliable diagnoses. Features extracted from a low-quality ECG signal can result in misinterpretations and potentially lead to misdiagnosis. Frequent disturbances such as electrical interference, muscle movement, and motion artifacts can lead to undesired noises in the signal. The ECG signal can be partially or completely corrupted by these noises, leading to inaccurate ECG delineation and diagnosis decisions based on these delineated waves. Machine learning (ML) and deep learning (DL) techniques become essential for automatically identifying the corrupted ECG signal quality. In this work, we proposed an automated method for classifying ECG signals into “acceptable” and “unacceptable” categories using deep convolutional neural networks. Two additional classifiers were added to the deep convolutional model to overcome the vanishing gradient problem. We trained and evaluated the model using a publicly available CinC 2011 dataset. The proposed model demonstrated 95.48% accuracy, 98.60% sensitivity, 82.35% specificity, and 97.24% F1 score. The performance of the proposed model was compared with existing state-of-the-art models in the literature as well the well-established deep learning architectures such as ResNet18, GoogleNet, and AlexNet. Our proposed model outperformed most of the models in the literature.