News and Events

Seminar by T. V. Sreenivas on Stochastic approaches to Music/Speech modeling

Title: "Stochastic approaches to Music/Speech modeling" by T.V. Sreenivas (Indian Institute of Science, Bangalore, India)

When and where? Tuesday October 23rd, 3:30pm in room 55.309

Abstract: Among the most prolific of signals that we deal with are speech and music, one being information rich and the other emotion (feelings) rich, along with sharing some of the characteristics between each other. Both types of signals are highly dynamic in nature, exhibiting a lot of variability due to individual characteristics of expression and style, in spite of underlying structural conventions. Stochastic models have been very successful in representing such variability in the signal patterns, along with structural variability also (as seen in speech models). Indian art-music (classical) is considered very structured and also practiced with high rigor, along with certain freedom for individual artistic expression. We examine the stochastic approaches in the literature to analyze Indian art-music and present our approach to estimate /shadja/, /swara/ and /rAga/, in an unsupervised manner. Through these models we draw parallels between the structure of speech and music signals and aim to explore the cognitive differences in the learning of speech and music.

16 Oct 2012 - 08:52 | view
Melody Extraction vamp plug-in available online!

We are very pleased to announce the launch of the MELODIA - Melody Extraction vamp plug-in during the ISMIR 2012 Conference which will take place in Porto, Portugal, October 8th-12th, 2012.

This plug-in implements our melody extraction algorithm which obtained good results in last year's MIREX Audio Melody Extraction campaign. Full details of the algorithm are available in:

J. Salamon and E. Gómez, "Melody Extraction from Polyphonic Music Signals using Pitch Contour Characteristics", IEEE Transactions on Audio, Speech and Language Processing, 20(6):1759-1770, Aug. 2012.

The plug-in is available online for free download (non-commercial purposes). A slightly less formal description of the algorithm, including graphs and audio examples is provided by the author Justin Salamon and can be found here.

In addition to benchmarking new algorithms against MELODIA, we hope it will serve the research community for research problems which could benefit from a predominant F0 estimator (e.g. query by humming, version identification, motif discovery and analysis, automatic transcription, source separation, etc.). We are very interested in receiving feedback from the research community, please let us know what you think!

Hope you enjoy the MELODIA experience!

5 Oct 2012 - 17:40 | view
Voctro Labs collaborates in the new "El Plan B de Ballantine's"

Voctro Labs, spinoff of the MTG, provides singing voice synthesis technology for the campaign "El Plan B de Ballantine's". The popular band "La Oreja de Van Gogh" composed a new lyric-less song (just music and melody) and they are inviting all their fans to participate in the creative process of composing the lyrics for this song. Fans can listen to the new song and compose the lyrics on the campaign's web site. Voctro Labs created a new female Vocaloid voice for this project, which permits fans to figure out how it would sound if it was sung by a real singer.

The PlanB web site is online since October 1st, drawing the attention of the media and Vocaloid fans worldwide. Also, at the end of the year, La Oreja de Van Gogh will choose their favorite lyrics sent by the fans and will use them for the final version of the song.

5 Oct 2012 - 11:49 | view
Seminar by Gautham Mysore on Non-negative Hidden Markov Modeling of Audio

When and where? Thursday, Oct 4, 2012, 3:30pm, 52.321

Host: Xavier Serra (MTG)

Title: Non-negative Hidden Markov Modeling of Audio

Abstract:
Non-negative spectrogram factorization techniques have become quite popular in the last decade as they are effective in modeling the spectral structure of audio. They have been extensively used for applications such as source separation and denoising. These techniques however fail to account for non-stationarity and temporal dynamics, which are two important properties of audio. In this talk, I will introduce the non-negative hidden Markov model (N-HMM) and the non-negative factorial hidden Markov model (N-FHMM) to model single sound sources and sound mixtures respectively. They jointly model the spectral structure and temporal dynamics of sound sources, while accounting for non-stationarity. I will also discuss the application of these models to various applications such as source separation, denoising, and content based audio processing, showing why they yield improved performance when compared to non-negative spectrogram factorization techniques.

2 Oct 2012 - 16:49 | view
Big participation of MTG researchers at ISMIR 2012

12 papers discussing research done at the MTG are being presented at the 13th International Society for Music Information
Retrieval Conference, that takes place in Porto from October 8th to the 12th 2012. These are:

 

2 Oct 2012 - 09:07 | view
New graduate students at the MTG

This academic year 2012-2013, 19 students are starting the SMC Master, 7 students are doing the second year of the Master, and 4 students are starting the PhD.

First year Master students: Robert Clouth (UK), Jakab Pilaszanovich (Hungary), Jose Javier Valero (Spain), Filippo Morelli (Italy), Úrbez Capablo Riazuelo (Spain), Jie Chen (China), António Humberto Sá Pinto (Portugal), Gabriel Meseguer (Spain), Dara Dabiri (Iran), Ken Chen (China), William Thomas Marley (Irland), Oriol Romaní Picas (Spain), Adrià Casas Gimeno (Spain), Toshinao Ruike (Japan), Roger Rios Bubiras (Spain), Aram Estiu Graugés (Spain), Raquel Victoria Baez Fernández (Spain), Macià Forteza Crespí (Spain), Nadine Kroher (Austria).

Second year Master students: Aluizio Neto (Brazil), Enric Giné  (Spain), Felipe L. Navarro (Spain), Hector Floria (Spain), Jakue Lopez (Spain), Varun Jewalikar (India), Hector Parra (Spain).

PhD students: Juanjo Bosch (Spain), Sankalp Gulati (India), Ajay Srinivasamurthy  (India), Sergio Giraldo (Colombia).

23 Sep 2012 - 17:37 | view
UPF Social Council awards

The resolution of the 2012 Social Council Awards was published yesterday in the course of the opening ceremony of the Academic Course 2012-2013. We are very proud to announce that the following MTG members were awarded:

  • Award in recognition to Quality and innovation in teaching: Emilia Gómez and Justin Salamon
  • Award in recognition to the track-record in Technology Transfer: Jordi Bonada

Congratulations to the winners!!!

 

18 Sep 2012 - 18:33 | view
Gert Lanckriet, from UCSD, visits the MTG

He will meet MTG researchers and give a seminar at our department. Everyone is invited!


***

Gert Lanckriet, University of California, San Diego 

Wednesday, Sep 26, 2012, 12pm, 52.421

Host: Emilia Gómez (MTG)>

Title: Music Recommendation with Multi-Modal Metric Learning to Rank

<Abstract:
A revolution in music production, distribution and consumption made millions of songs available to virtually anyone on the planet, through the Internet. To allow users to retrieve the desired content from this nearly infinite pool of possibilities, algorithms for automatic music indexing and recommendation are a must.

In this talk, I will discuss two aspects of automated content-based music analysis for music search and recommendation: i) automated music tagging for semantic retrieval, and ii) a query-by-example paradigm for content-based music recommendation, wherein a user queries the system by providing a song, and the system responds with a list of relevant or similar song recommendations (e.g., playlist generation for online radio).

Query-by-example applications ultimately depend on the notion of similarity between items to produce high-quality results. Current state-of-the-art systems employ collaborative filter methods to represent musical items, effectively comparing items in terms of their constituent users. While collaborative filter techniques perform well when historical data is available for each item, their reliance on historical data impedes performance on novel or unpopular items. To combat this problem, we rely on content-based similarity, which naturally extends to novel items, but is typically out-performed by collaborative filter methods. In this talk, I will present a method for optimizing content-based similarity by learning from a sample of collaborative filter data. Finally, I will discuss how such algorithms may be adapted to improve recommendations if a variety of information besides musical content is available as well (e.g., music video clips, web documents and/or art work describing musical artists).

Biography:
Gert Lanckriet received a Master's degree in Electrical Engineering from the Katholieke Universiteit Leuven, Leuven, Belgium, in 2000 and the M.S. and Ph.D. degrees in Electrical Engineering and Computer Science from the University of California, Berkeley in 2001 respectively 2005. In 2005, he joined the Department of Electrical and Computer Engineering at the University of California, San Diego, where he heads the Computer Audition Lab. He was awarded the SIAM Optimization Prize in 2008 and is the recipient of a Hellman Fellowship, an IBM Faculty Award, an NSF CAREER Award and an Alfred P. Sloan Foundation Research Fellowship. In 2011, MIT Technology Review named him one of the 35 top young technology innovators in the world (TR35). His research focuses on the interplay of convex optimization, machine learning and applied statistics, with applications in computer audition and music information retrieval

18 Sep 2012 - 17:38 | view
VoctroLabs awarded by SIMO

Voctro Labs, our third spin-off company, was one of 14 start-ups winners of the 7th edition of 'Vivero 2012 award', organized by SIMO network. This award will allow Voctro to have a stand at the fair of the same name (SIMO) which takes place from 25 to 27 September in Madrid IFEMA Exhibition.

Voctro Labs has been recognized for their work in the development of audio and voice processing technologies applied to the entertainment sector. With this appointment Voctro Labs can earn 9,000 euros (a Special Prize awarded by Fundetec).

12 Sep 2012 - 09:10 | view
The MTG takes part in "Programa Professors i Ciència" (Fundació Catalunya Caixa)

The MTG collaborates in the "Programa Professors i Ciència" (Teachers & Science program), funded by Fundació Catalunya Caixa.

The program offers high-school teachers the opportunity of taking part in scientific specialization courses at research centers in Catalonia. In this way, the program aims to bring research closer to educational institutions at the secondary level. The MTG organizes a course on sound & nature that will take place in November.

 

7 Sep 2012 - 11:44 | view
intranet