Master Thesis Report v49

ii iii Acknowledgements I would first like to thank Prof. Xavier Serra for giving me the opportunity to join the SMC Master's program at the Pompeu Fabra University, and the whole Music Technology Group for providing an excellent intellectual framework for the completion of this thesis, specially the members and SMC students which have somehow influenced the development of this work. Special thanks to Perfecto Herrera and Ferdinand Fuhrmann for all their ideas, recommendations, revisions, and for the provision of the code and original dataset that I used as a basis for the instrument recognition in this work. Many thanks to thank Ricard Marxer for the clarifications, as well as Yamaha for the opportunity they provided me to work with their separation algorithm. Special thanks as well to Jordi Janer for his supervision, clarifications, suggestions, and patience. I would like to specially thank my family for all their love and support throughout the years, without which I could not have written this thesis. Thanks as well to friends and everyone else that supported me during this year, such as Andrea with her computing hardware (and patience). Many thanks as well to "La Caixa" Fellowship program for their confidence and their financial support during this year. iv Abstract Due to the increasing amount of digital music available, there is a clear need of a proper organization and effective retrieval. Automatic instrument recognition techniques are useful for satisfying such needs, by labeling music pieces with their instrumentation, but also as support to the extraction of other semantic information such as the genre. Source separation has also recently been applied to facilitate the analysis of musical data, as well as to other applications such as karaoke or post production. In contrast with the huge need for both algorithms, the results obtained so far show that there is still much room for improvement. The main purpose of this thesis is to find synergies between instrument recognition and source separation algorithms in two different tasks: 1) the separation of a target instrument from the accompaniment, and 2) the automatic labeling of songs with the predominant music instruments. Several combination strategies are presented, aimed at overcoming some of the limitations of current state-of-the-art algorithms. In the first task, instrument recognition is used to detect the presence of the target instrument in order to apply or bypass the separation algorithms. In the second task, source separation is …