Title: An approach for melody extraction from polyphonic audio: Using perceptual principles and melodic smoothness
Abstract: In this research work, the problem of melody extraction from polyphonic audio is addressed. A multi-stage approach is followed, inspired on principles from perceptual theory and musical practice. Physiological models and perceptual cues of sound organization are incorporated into the method, mimicing the behavior of the human auditory system to some extent. Moreover, musicological principles are applied, in order to support the identification of the musical notes that convey the main melodic line. The system comprises three main modules, where a number of rule-based procedures are proposed: (i) Pitch detection, where an auditory model-based pitch detector is employed for selecting multiple pitches in each analysis frame; (ii) determination of musical notes (with precise temporal boundaries and pitches); and (iii) identification of melodic notes, based on two core assumptions that we designate as the salience principle and the melodic smoothness principle. Experimental results were conducted, showing that the method performs satisfactorily under the specified assumptions, namely when the notes comprising the melody are, in general, more intense than the accompanying instruments. However, additional difficulties are encountered in song excerpts where the intensity of the melody in comparison to the surrounding accompaniment is not so favorable.