News and Events

This month, January 2008, has started a new european project within the new FP7 of the EU in which the MTG will carry out research on voice transformations and auditory stream processing. The project is called SAME, Sound And Music for Everyone Everyday, and the partners are: Università degli Studi di Genova (coordinator), Nokia Research Center, Kungliga Tekniska Högskolan, Pompeu Fabra University, Helsinki University of Technology, Institut de Recherche et Coordination Acoustique/Musique.

The SAME project falls in the Networked Media Objective, and will develop (i) an end-to-end research platform to exploit new music and multimedia metadata, and context data usage from end users’ mobile devices enabling creation, manipulation, rendering of media by non-professional users, and (ii) new creative forms of interactive, immersive, and very high quality media (including 3D audio and augmented reality), to enable new forms of experience for individual users or user communities, on context-aware, active mobile music listening.

The MTG research within the SAME project will mainly relate to techniques for voice transformations and auditory stream processing. For both singing and speech, voice transformations research will focus on:

• Natural character transformations based on high-level descriptions (such as gender change, age change, deeper voice, creaky voice…), including expressive/emotional descriptors;
• Voice impersonation based on high-level descriptors: transform an input voice into a target one by comparing their descriptors obtained from analysis, and inferred behavioral models.

• Enhancement and combination of state of the art algorithms suitable for key instrument characterization based on predominant pitch estimation, multipitch estimation, and aural localization in a stereo mix;
• Identification of relevant transformations (including voice transformations);
• Research and implementation of identified transformations;
• Quality assessment by means of perceptual experiments and quantitative evaluation

Juntament amb el Guillamino el Reactable va posar la música i el control de les llums a les 12 campanades de la nit de Cap d'Any de TV3 a la Torre Agbar de Barcelona.

Video

Noticia a la Web de la UPF

SMCNetwork.org is a new web portal hosted by the MTG aiming at becoming a useful information and discussion forum for the international Sound and Music Computing research community. Building on the outcome of the European project S2S2 it already includes a new version of the SMC Roadmap and some relevant resources.

SMCNetwork.org

Perfecto Herrera will give a talk at the Max Planck Institute for Human Cognitive and Brain Science in Leipzig, Germany, entitled "Automatic music content description: from signals to symbols." The talk will be on Wednesday 19th of December at 3pm.

The talk will present and discuss some of the current problems and methodologies for extracting, from an audio music file, "objective" representations of its musical and sonic content. The extracted descriptors belong to different levels of abstraction and help to characterize different musical facets. Even though they do not provide a musical "transcription", several music classification problems (e.g., genre classification, version detection, mood classification, key and mode detection, etc.) can be addressed with a high degree of success. Therefore, commercial applications such as music recommenders, music+fitness systems or radio airplay monitoring, can be built on top of them. Even though this apparent success, it is clear to music technologists that knowledge about how human brains process musical stimuli, incomplete as it is at this point, may help to guide the inception of improved descriptors and music content processing algorithms. At the same time, some of our technologies can be useful in the preparation and analysis of music psychology and neuroscience experiments.

More information on the Institute for Human Cognitive and Brain Science: MPI CBS

The MTG has obtained a grant from the Google Research Awards program to support the Freesound project. With the award of $50.000 US dollars the MTG will be able to make some major improvements to Freesound.

The MTG believes that the potential and future impact of Freesound is enormous and that it can become the reference site for sound sharing in the very near future. With this award we want to (1) increase the rate at which new sounds are being added, (2) improve the software so that the site can scale up properly, and (3) add functionality to further promote its use in research and creative activities worldwide. This award will allow to do all this while maintaining the current model of a free and collaborative site based on a creative commons license.

This is an award for the whole Freesound community, so congratulations to everyone!!!.

Joachim Neumann will present the latest MTG technology within the scope of the VARIAZIONI project presentation at the AXMEDIS Conference in Barcelona, 29 November 2007. The Variazioni project aims to centralize music and audio collections from music schools and conservatories around Europe, and make them accessible for the wider public through a webportal with social web and semantic media technologies. The technologies MTG will present includes automatic tagging of songs & automated music analysis, segmentation & similarity.

More information: AXMEDIS Conference.

Hendrik Purwins and Xavier Serra are co-organizers of the workshop "Music, Brain and Cognition" at NIPS 2007 on December 7th and 8th in Whistler, Canada.

This workshop will span topics from signal processing and musical structure to the cognition of music and sound. It will include the following paper presentations by MTG researchers:

• R.Annies, E. Martinez, K. Adiloglu, H. Purwins and K. Obermayer. "Comparison of Biologically Inspired Classification Schemes for Everyday Sounds"
• M.Coath, S. L. Denham, L. Smith, H. Honing, A. Hazan, P. Holonowicz and H. Purwins. "An Auditory Model for the Detection of Perceptual Onsets and Beat Tracking in Singing".
• R.Marxer, P. Holonowicz and H. Purwins. "Dynamical Hierarchical Self-Organization of Harmonic, Motivic, and Pitch Categories"
• A. Hazan, P. Brossier, R. Marxer and H. Purwins. "What/When Causal Expectation Modelling in Monophonic Pitched and Percussive Audio"

Xavier Serra will give, on thursday 8th of November at 3pm in room 102 of the França building, a seminar entitled "State of the Art and Future Directions in Musical Sound Synthesis."

In this talk Xavier will first place the sound synthesis topic within its research context, then he will highlight some of the current trends, and finally he will attempt to identify some challenges for the future.

More information on Xavier's talk: Research Seminar