Masters Thesis

Sequential Multi-pitch Tracking

We recently developed a state-of-the-art multipitch-detection algorithm [1] that works on a frame-by-frame basis. However, it is expected that the algorithms performance can be improved by tracking the appearance/disappearance of pitches (i.e. tones) over multiple frames. Thus, the aim of this thesis to use the frame-by-frame estimates of [1] as pre-processing algorithm for a multi-object tracking algorithm, e.g. based on belief propagation (BP) and the sum-product-algorithm (SPA) [2].

Your Tasks

  • Literature research on (sequential) multi-pitch detection algorithms.
  • Implement a multi-pitch tracking algorithm based on SPA in python or matlab (reference code from [2] is available).
  • Evaluate the performance, e.g. on the Bach 10 dataset.
  • Write your thesis.
  • (Optional, if the results are promising) write up your results in a short conference paper (e.g. ISMIR conference).

Your Profile

  • Familiar with (statistical) signal processing.
  • Inclined to do theoretical/simulation based work.
  • Some experience with python or matlab is beneficial.

References

  1. J. Möderl, F. Pernkopf, K. Witrisal, and E. Leitinger, "Variational Inference of Structured Line Spectra Exploiting Group-Sparsity." arXiv preprint, 2023, doi: 10.48550/arXiv.2303.03017
  2. X. Li, E. Leitinger, A. Venus and F. Tufvesson, "Sequential detection and estimation of multipath channel parameters using belief propagation," IEEE Transactions on Wireless Communications, 2022, doi: 10.1109/TWC.2022.3165856
Contact
image/svg+xml

Jakob Möderl

Erik Leitinger (SPSC)