Google Suche
Prof. Dr.-Ing. Gerald Schuller
Fachgebietsleiter
+49 3677 69-2756 | Fax: +49 3677 69-2888
E-Mail senden
Kirchhoffbau, Raum K 3014
Sprechstunde nach Vereinbarung
Sekretariat
Marina Bondarev
Kirchhoffbau, Raum 3007
49 3677 69-2890 | Fax: +49 3677 69-2888
E-Mail senden
Besucheradresse
Gustav-Kirchhoff-Straße 1, Raum K 3007 (3. Etage)
98693 Ilmenau
Postanschrift
Technische Universität Ilmenau
Fakultät für Elektrotechnik und Informationstechnik
Fachgebiet Angewandte Mediensysteme
PF 100565
98684 Ilmenau
Deep learning has some big successes generating artificial images, for instance of faces, using "Generative Adverserial Networks" or "Variational Autoencoders" (VAE), so-called "Deep Fakes".
A simple example program for handwritten images can be found here: https://github.com/kvfrans/variational-autoencoder
The goal of this project is to use the VAE Network to generate musical instrument sounds, similar to "Nsynth". https://magenta.tensorflow.org/nsynth-instrument.
For this, the VAE is trained on musical instrument sounds from the IDMT Musical Instruments Database. A suitable training set has to be chosen and tested, and a good set of hyper-parameters (the dimensionality of the network) has to be found. Then a psycho-acoustic similarity measure based on our psycho-acoustic pre- and post-filters has the be used and tested for the VAE network (for the "generation loss"), and compared to other similarity measures.
Supervision: Prof. Dr.-Ing. G. Schuller
Backback