NTR webinar: Single-channel source separation for audio data

NTR organizes scientific webinars about neural networks and invites speakers from all over the world to lead them. 

On January 19 Olga Slizovskaia, R&D Engineer, Voctro Labs S.L., Barcelona, Spain, held a technical Zoom webinar on Single-channel source separation for audio data: modern methods and applications. 

What the webinar covered: 

  • Single-channel source separation in audio data, starting with an overview of modern source separation methods describing their applications in music and speech. 
  • The different input data representations, including reviewing the pros and cons of using spectrograms and raw audio. 
  • A detailed framework of modern supervised and self-supervised methods, both in the classical audio-only context, as well as in a scope of score- and video-assisted source separation. 
  • Outline solutions for cases of an unknown number of sources in the data and discuss quantitative metrics for quality assessment.

Materials available:

  • Webinar recording on YouTube (in Russian)

Moderator and contact:

NTR CEO Nick Mikhailovsky: nickm@ntrlab.com.

Leave a Reply

Your email address will not be published. Required fields are marked *