Cardiovascular diseases continue to be a major cause of morbidity and mortality worldwide. Electrocardiograms (ECGs) play a crucial role in the diagnosis and monitoring of various cardiac ailments. In this study, we present a novel approach for the classification of cardiovascular ECGs using Maximal Overlap Discrete Wavelet Packet Transform (MODWPT)-based feature extraction. The dataset comprises 1200 records of segmented ECG signals from four distinct ailments: MIT-BIH Arrhythmia, BIDMC Congestive Heart Failure, MIT-BIH Atrial Fibrillation, and MIT-BIH Normal Sinus Rhythm, sourced from the MIT-BIH physio-net database. We preprocess the ECG signals with bandpass filters and normalize them using gain specific to each database. Subsequently, we apply MODWPT to extract 54 informative features, which serve as inputs to the classification models. To comprehensively assess the model performances, we evaluate five popular classifiers: Logistic Regression, Support Vector Machine, Gradient Boosting Classifier, Random Forest Classifier, and K-Nearest Neighbours. The results demonstrate remarkable accuracy and robustness in classifying ECG signals into the four distinct ailments. Gradient Boosting Classifier and Random Forest Classifier achieved perfect precision, recall, F1 Score, and ROC AUC, highlighting their exceptional performance. SVM and K-Nearest Neighbours also exhibit high accuracy and classification metrics, while Logistic Regression performs well but shows a relatively lower ROC AUC.