This research project uses intelligent matching with AI to identify musical works for efficient and legally compliant digital licensing. In the age of digitalisation, traditional reporting and recording of music usage is no longer sufficient.

Industry
Intellectual Property

Services
Intelligent audio matching
Management of plant information
Demonstrator development


Quick Facts

81 %
Cover song recognition
3
Duration of the R&D project in years
1.66 billion €
Sales of digital music in DE 2022
80,3 %
Market share of digital music in DE 2022

The challenge

Traditional reporting and recording of music usage can no longer keep pace with the demands of rapidly advancing digitalisation. This research project therefore aims to overcome the complexity of digital music usage with intelligent, AI-supported matching. Our approach focusses on the identification of musical works in different versions. By using AI and usage tracking in a smart contract, we create an innovative process that enables legally compliant and flexible digital licensing. Our research aims to support the music industry with an efficient and accurate system for licensing and billing.


The solution

Matching strategy 1: Fingerprint Matcher
This strategy uses a Fast Fourier Transformation to create fingerprints of music files (on average 100,000 fingerprints per file). This method can recognise precisely known pieces of music and also individual pieces in composite titles. However, it cannot recognise the relationship between album and concert versions, nor can it assign cover versions to the original. It also requires approx. 1 MB of memory per minute of audio.
Matching strategy 2: CQTNet + Milvus
This strategy uses a Convolutional Neural Network (CNN) to classify feature vectors created using Constant-Q Transform (CQT). This method can precisely recognise known pieces of music and assign both album and concert versions as well as cover versions to the original. Although it cannot recognise individual pieces in composite works, it only requires approx. 2.5 KB of memory per piece of music.
Usage Tracking
Usage tracking was developed by research partner TUM using a smart contract based on a Hyperledger blockchain.

Technologies used

About the research partners

With over 52,500 students, the Technical University of Munich (TUM for short) is the largest technical university in Germany in terms of student numbers. It is based in Munich. It was one of the first three universities to be included in the ‘Institutional Strategy’ funding line as part of the Excellence Initiative in 2006. It successfully defended its title as a University of Excellence in 2012 and 2019. It is part of the Bavarian Elite Network and enjoys a high academic reputation.

Research programme

This work was funded by the Bavarian State Ministry of Economic Affairs, Regional Development and Energy as part of the Bavarian Joint Funding Programme (BayVFP) – Digitisation funding line – Information and Communication Technology funding area.


Do you have any questions? We will be happy to help you!

Do you have a project that you would like to talk to us about? You are welcome to contact us at any time. We look forward to getting to know you!

Marc Ponschab
Marc Ponschab Head of Technology Lab / R&D