Single Channel Source Separation applied to Polyphonic Music

Project Type: Student Project
Student: Clara Maria Hollomey
Mentor: Robert Peharz

Short Description

In this project we want to apply single channel source separation (SCSS) techniques to the problem of separating musical instrument sources from a recorded piece of music. For this task the factorial max-Vector Quantization (max-VQ) and the factorial Sparse Coder (SC) models should be used, which have been successfully applied to human speech.

Your Tasks

  • Getting an overview of SCSS, especially with factorial max-VQ and factorial SC
  • Apply factorial max-VQ and factorial SC to synthetic music
  • Identify specific challenges which occur in connection with music
  • Possibly modify the SCSS systems
  • Possibly apply SCSS systems to real-world music
  • Write a report (around 10 pages)

Your Profile/Requirements

This project is suited for Master students in Telematics, Electrical Engineering, Audio Engineering, Computer Science and Software Development.


Robert Peharz ( or 0316/873 4482)


[1] S. Roweis, “One microphone source separation,” in Neural Information Processing Systems, 2001, pp. 793–799.

[2] ——, “Factorial models and refiltering for speech separation and denoising,” in EUROSPEECH, 2003, pp. 1009–1012.

[3] P. Smaragdis, B. Raj, and M. Shashanka, “Supervised and semi-supervised separation of sounds from single-channel mixtures,” 2007.