Deep-Music-Dream: A New Artificial Intelligence Approach to Computer-Aided Composition
Lead Research Organisation:
Plymouth University
Department Name: Sch of Humanities & Performing Arts
Abstract
Adapting image manipulation techniques such as those used by Google DeepDream into a new artificial intelligence system for musical composition, through analysis & reconstruction of spectrograms. A primary objective for this will involve the use of deep learning systems, specifically neural networks - interconnected clusters of code which interact with each other in a similar way to neurons in the human brain, and are especially useful for pattern recognition.
Organisations
People |
ORCID iD |
Eduardo Miranda (Primary Supervisor) | |
Samuel Pearce-Davies (Student) |
Description | A dataset of spectrograms (pictures of sound) have been generated from Magenta's NSynth dataset. Appropriate spectrogram generation parameters have been determined for better computational efficiency without substantial loss of quality. Machine learning using spectrograms has been performed in a novel environment (Max/MSP) to produce rudimentary original instrumental notes. |
Exploitation Route | Broadly to be determined as the award is still active, but the generated spectrogram dataset may be distributed among the community for additional AI training at various levels. |
Sectors | Digital/Communication/Information Technologies (including Software),Other |