Data augmentation techniques for cover song identification

• Grant number: 19/08653-0
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: July 01, 2019
• End date: August 31, 2020
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: Diego Furtado Silva
• Grantee: Ricardo Szram Filho
• Host Institution: Centro de Ciências Exatas e de Tecnologia (CCET). Universidade Federal de São Carlos (UFSCAR). São Carlos , SP, Brazil

Abstract

Music has been part of the human being's daily life for centuries. With the evolution of society and its productive means, this art was increasingly becoming part of commercial activities. This is currently very evident thanks to streaming music platforms. One of the problems faced by these platforms is the attribution of copyright, especially when users are allowed to create their own content. The automatic recognition of cover songs is a part of the actions to improve this process. The major problem faced by the tools for this task is the fact that cover songs have severe variations from the original recording, such as changes in tempo and timbre. Thus, the comparison of the similarity between the feature vectors that describe the songs is impaired. The purpose of this research is to investigate audio distortion techniques and generative adversary neural networks to create synthetic data in order to improve the effectiveness of state-of-the-art algorithms for automatic recognition of cover songs.