Descripció del projecte
In today’s digital music market, ensuring industry quality standard of the audio masters and preventing copy rights infringement are two of the most challenging technical problems. In the digital music supply chain industry, failing to detect this issues can lead to important risks, such as legal claims from the rightful owners of the content or legal penalties from music services due to bad quality content. As a result, SonoSuite is constantly working on researching and prototyping solutions that can tackle this problem. Concurrently and thanks to the omnipresence of digital tools, musicians have adopted sampling as part of their creative process by using musical samples of other works to create a new composition. The use of samples needs to be allowed by the copy rights owner by proving a license of use. Detecting automatically these uses to ensure the appropriate licenses are in order before distribution would minimise risks for the company, clients and artists.
The main goal of this project is the research and development of sample detection algorithms using deep learning technologies together with music information retrieval (MIR) content-based systems and digital signal processing techniques for the automatic extraction of features and data from the audio. As a secondary goal, manual and automated methods will be proposed for annotation of musical data. Furthermore, databases with annotated data for use in deep learning models training will be developed.
In order to contribute with new ideas and approaches for sample detection it will be needed to study the state of the art of some other MIR fields such as: classification of sound events, classification of music by genre, fingerprinting technologies for music identification ,and cover detection methodologies.
