Pablo González Carrizo

Sound and Image Engineering in Telecommunication

Polytechnic School. University of Alicante.

In this project, we will focus on one concrete area of the music transcription: automatic chord estimation (ACE). We will analyze the main subsystems inside an automatic chord estimation system, and how, modifying theirs parameters, changes the result of the estimation.
The analyzed subsystems have been divided in some categories depending on the part of the system where they are executed: signal adaptation, feature extraction, prefiltering, classification and postfiltering. We have analyzed the effect on the system of harmonic- percussive separation, signal filtering or beat synchronization. However, as we have seen, the more important parts of the system were the training dataset and the classifier used. We have used two different kind of training datasets. One of them comes from ideal theoretical models. The other one, comes from real samples taken from musical instruments. Three dif


Technical report (in spanish): Pablo Gonzalez.pdf

Source code: