A real-time music-scene-description system: predominant-FO estimation for detecting melody and bass lines in real-world audio signals

摘要

In this paper, we describe the concept of music scene description and address the problem of detecting melody and bass lines in real-world audio signals containing the sounds of various instruments. Most previous pitch-estimation methods have had difficulty dealing with such complex music signals because these methods were designed to deal with mixtures of only a few sounds. To enable estimation of the fundamental frequency (F0) of the melody and bass lines, we propose a predominant-F0 estimation method called PreFEst that does not rely on the unreliable fundamental component and obtains the most predominant F0 supported by harmonics within an intentionally limited frequency range. This method estimates the relative dominance of every possible F0 (represented as a probability density function of the F0) by using MAP (maximum a posteriori probability) estimation and considers the F0's temporal continuity by using a multiple-agent architecture. Experimental results with a set of ten music excerpts from compact-disc recordings showed that a real-time system implementing this method was able to detect melody and bass lines about 80% of the time these existed. (C) 2004 Elsevier B.V. All rights reserved.