Category Uncategorised

On 7-13 & 22-27 May, a number of AIM researchers will participate at the 2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP 2022). ICASSP is the leading conference in the field of signal processing and the flagship event of the IEEE Signal Processing Society. The below papers presented at ICASSP 2022 are authored or co-authored by AIM members:

• Direct design of biquad filter cascades with deep learning by sampling random polynomials, by Joseph T. Colonel, Christian J. Steinmetz, Marcus Michelen, Joshua D. Reiss
• Improving lyrics alignment through joint pitch detection, by Jiawen Huang, Emmanouil Benetos, Sebastian Ewert
• Learning music audio representations via weak language supervision, by Ilaria Manco, Emmanouil Benetos, Elio Quinton, George Fazekas

See you all at ICASSP!

The AIM 2020 cohort will be organising the Digital Music Research Network (DMRN) 2021 workshop on 21st December 2021. The workshop is hosted at QMUL on an annual basis by the Centre for Digital Music and aims to promote research in Digital Music by bringing together researchers from academia and industry in electronics engineering, computer science, and music.

The workshop will include 2 keynote talks by Dr. Sophie Scott and  Dr. Gus Xia  along with a neural synthesis hackathon and poster presentations. For more details, please refer to the relevant website.

As a satellite event to DMRN+16, AIM researchers have organised a hackathon around the theme of neural audio synthesis, a growing field which focuses on producing audio with neural networks.

The hackathon is organised by SIGNAS (the Special Interest Group on Neural Audio Synthesis), a group formed in response to the growing contingent of researchers at C4DM interested in topics at the intersection of sound perception, synthesis, and deep learning.

The neural audio synthesis hackathon (NASH) will take place on 18th-19th December, with the winning hacks to be presented at a talk at DMRN+16 on the 21st December. At the time of writing nearly 50 participants have registered, and this number continues to grow as the event approaches. The event aims to encourage cross-disciplinary collaboration in neural audio synthesis by encouraging the development of new techniques, tools, and interfaces, with a particular focus on creative musical applications. Projects are particularly encouraged under the following topics:

• Interfaces and instruments
• Novel techniques and models
• Synthesis control
• Creative applications

More information and registration details can be found on the hackathon website.

In advance of the hackathon, SIGNAS have also organised a neural audio synthesis workshop on 12th December at 4pm UTC, which will feature tutorials from 5 exciting researchers and artists working with deep learning and audio synthesis, including AIM’s own Ilaria Manco. The event is free and will be hosted online, with tickets and details about talks available at this link.