From June 7 to 17, I am visiting Rémi Gribonval and other colleagues at L’IRISA (Institut de recherche en informatique et systèmes aléatoires) in Rennes, France, where I am on a scientific mission supported by the French Ambassador to Denmark. My goals this first visit are twofold:
- implement cyclic MP (CMP) within the MPTK framework;
- empirically study its application to real audio and music signals.
Before this visit, CMP only existed in the MATLAB backwaters of few esoterics;
and there, it is only usable for small dictionaries, and only applied to short signals.
We have now finished implementing CMP in MPTK, which required only simple modifications — no doubt due to the well-thought architecture of the MPTK library.
And now begins the more difficult process of determining the cost-benefit trade-offs for CMP and other variants.
The plot below shows the first result.