Final Report Summary - STAMUS (Statistical models for musical signal processing)
The objective of the STAMUS project was the development of innovative technologies for enabling access to the immense amount of audio music collections that are available on the Web. Although there have been considerable advances in music storage, distribution, indexation and many other directions in the last decades, there are still some bottlenecks for the analysis and extraction of content information. Music audio signals are very complex, both because of the intrinsic nature of audio, and because of the information they convey. The STAMUS project sought to build efficient models that allow the automatic extraction of musical content information from audio signals.
The STAMUS project has addressed two important limitations of state-of-the-art Music Information Retrieval (MIR) systems for estimating musical content from audio signals. At the statistical modelling level, it has developed innovative models that are able to handle the semantic complexity of music. In particular it has developed algorithms for music analysis that are able to take into account the music structure at several time scales. For instance the highest-level expression of the structure (segmentation of songs into verse/chorus, etc.) is related to musically lower-level organization such as the chord progression. The developed models for music analysis allow modelling this hierarchical structure by favouring strong similarity between the chord progression of two repetitions of semantically same segments (e.g. two verse in a song). At the signal representation level and at the signal features level, STAMUS project has developed approaches to take into account the ‘adverse environment’ signal features are necessarily corrupted by (e.g. background noise). New well-structured representations have been proposed, which allow providing access to higher-level information about the audio signal, taking into account the high variability of audio and the effect of the adverse environment. For instance the various components that are usually superimposed in music signal (tonal components, transients, background noise, etc.) can be separated, so that extraction of semantically meaningful music content information (such as chords, key, beats, etc.) is easier. Finally, the project has supplied new annotated audio corpora, which are necessary resources for the development and benchmarking/evaluation of MIR systems.
The work developed within STAMUS project has made progress towards the analysis of semantically complex music signals and the development of multimodal approaches for music analysis. The dissemination of these results has been ensured via academic publications and presentations in conferences that gather the main academic and industry stakeholders of the music domain. Music industry bodies and professionals have access to the results of the STAMUS project, which can be integrated in their music software. An overview of the project with a detailed bibliography can be found at http://stamus-project.blogspot.fr/. The STAMUS project has contributed to the continuous effort in the Music Information Retrieval community toward developing methods to enable or improve multimedia retrieval, which is one of the main technical and socio-economic challenges of today.
The STAMUS project has addressed two important limitations of state-of-the-art Music Information Retrieval (MIR) systems for estimating musical content from audio signals. At the statistical modelling level, it has developed innovative models that are able to handle the semantic complexity of music. In particular it has developed algorithms for music analysis that are able to take into account the music structure at several time scales. For instance the highest-level expression of the structure (segmentation of songs into verse/chorus, etc.) is related to musically lower-level organization such as the chord progression. The developed models for music analysis allow modelling this hierarchical structure by favouring strong similarity between the chord progression of two repetitions of semantically same segments (e.g. two verse in a song). At the signal representation level and at the signal features level, STAMUS project has developed approaches to take into account the ‘adverse environment’ signal features are necessarily corrupted by (e.g. background noise). New well-structured representations have been proposed, which allow providing access to higher-level information about the audio signal, taking into account the high variability of audio and the effect of the adverse environment. For instance the various components that are usually superimposed in music signal (tonal components, transients, background noise, etc.) can be separated, so that extraction of semantically meaningful music content information (such as chords, key, beats, etc.) is easier. Finally, the project has supplied new annotated audio corpora, which are necessary resources for the development and benchmarking/evaluation of MIR systems.
The work developed within STAMUS project has made progress towards the analysis of semantically complex music signals and the development of multimodal approaches for music analysis. The dissemination of these results has been ensured via academic publications and presentations in conferences that gather the main academic and industry stakeholders of the music domain. Music industry bodies and professionals have access to the results of the STAMUS project, which can be integrated in their music software. An overview of the project with a detailed bibliography can be found at http://stamus-project.blogspot.fr/. The STAMUS project has contributed to the continuous effort in the Music Information Retrieval community toward developing methods to enable or improve multimedia retrieval, which is one of the main technical and socio-economic challenges of today.