"This project lies in the interdisciplinary field of Music Information Retrieval (MIR). The objective of the project is the development of innovative technologies for enabling access to the immense amount of audio music collections that are available on the Web. Bringing together efforts from speech and signal processing, statistical modeling, musicology and machine learning, we seek to build efficient models that allow the automatic extraction of musical content information from audio signals. We will develop innovative statistical approaches capable of processing information on multiple semantic levels. We will develop a fully-probabilistic model that consists of a single, multi-faceted hierarchical musical structure that is jointly estimated from the music, and that is capable of directly describing the epistemic uncertainty that arises from the effects of the adverse environment in audio music. Concepts from the missing feature theory, which have been applied in robust automatic speech recognition technology, will be integrated to improve the robustness of the MIR algorithms.The objectives and methodologies match the interdisciplinary research profile of the experienced researcher. The high degree of expertise of the two host organizations, both in MIR and Signal and Speech Processing, makes this ambitious project highly feasible. It will also ensure that the experienced researcher will acquire high-level complementary skills that will be fruitful for her research career development in Europe.This interdisciplinary project is a very exciting challenge, and, as the work is disseminated, we expect that it will place European MIR research in a leading position for years to come."