Dataset for ECG Time-Domain Feature Analysis Based on Melodic Theory and Machine Learning

This dataset provides a collection of melodic and temporal features extracted from five well-established ECG databases in PhysioNet (NSRDB, STDB, Arrhythmia, SVDB, and VFDB). It contains 1,610 engineered features derived from RR intervals transformed using music theory principles, including distributional characteristics (mean pitch, standard deviation, duration, range) and time domain and melody features (complete sequences, direction parameters, spatial and location features). Specifically designed for machine learning applications in coronary heart disease (CHD) detection, this feature set enables novel analysis of cardiac rhythms through melodic representation, achieving effective classification performance (AUC up to 0.92) in early CHD warning tasks.