Vocal melody extraction using patch-based CNN

TL;DR

This paper introduces a patch-based CNN model for vocal melody extraction that uses a novel time-frequency representation, achieving efficient training and competitive accuracy in polyphonic music analysis.

Contribution

The paper presents a new CNN architecture and data representation for vocal melody extraction, inspired by object detection techniques in image processing.

Findings

Achieves high speed in melody extraction

Demonstrates competitive accuracy with limited labeled data

Effective in polyphonic music environments

Abstract

A patch-based convolutional neural network (CNN) model presented in this paper for vocal melody extraction in polyphonic music is inspired from object detection in image processing. The input of the model is a novel time-frequency representation which enhances the pitch contours and suppresses the harmonic components of a signal. This succinct data representation and the patch-based CNN model enable an efficient training process with limited labeled data. Experiments on various datasets show excellent speed and competitive accuracy comparing to other deep learning approaches.