Skip to content
  metiss  

Internship, Master, Sound source separation, 2007-2008

Document Actions
Conception and implementation in C++ of an incremental version of the sound-source directions estimation DEMIX algorithm.

The aim of the internship is to design an incremental version of DEMIX, a sound-source directions estimation algorithm designed in the METISS team of IRISA,
and implement it in C++, so as to incorporate it within an existing audio signal processing toolbox.
The current version of DEMIX estimates correctly both the number of sources and the source directions, but is limited to short signals because it is not incremental.

Audio signals are often the result of an acoustical or artificial mixture, in known or unknown conditions, of some sound sources.
To process these signals, we need efficient methods of source separation, to decompose these signals into simpler parts, either to index or recognise or transform the data.
In the underdetermined case, that is when the number of sources is higher than the number of observed channels, source separation involves solving three subproblems :
detecting the number of sources, estimating the mixing matrix, and finding an adequate inversion strategy to obtain the sources.
DEMIX is coded in MATLAB, and it automatically estimates both the number of sources and the directions of the sources, in the instantaneous and anechoic case.
Performances of DEMIX are validated and right, but the algorithm can only deal with short signals (around ten seconds).

So, the aim of this internship is both to :
- design a new version of the DEMIX algorithm, that has to be incremental so as to decrease the complexity of the algorithm and consequently enable the algorithm to deal with longer signals ;
- implement this new version in C++, so as to incorporate it within an existing audio signal processing toolbox.

keywords : computer music, audio signal processing, C++
Skills :
- Computer science : C++, Matlab, Object-oriented programming, algorithm
- Signal processing : Fourier Analysis, PCA
- Mathematics : Probability, Statistics, Linear algebra
Duration : 6 months
Location : Irisa, équipe METISS, Rennes France
Qualification : Master degree, Engineering school
Salary : according to qualifications
Contact :  
Simon ARBERET, tel : +332 99 84 74 40
Rémi GRIBONVAL, tel :  +332 99 84 25 06

mails : forname.lastname@irisa.fr

Created by sarberet
Last modified 29.07.2009 04:57 PM expired
« May 2012 »
Su Mo Tu We Th Fr Sa
    1 2 3 4 5
6 7 8 9 10 11 12
13 14 15 16 17 18 19
20 21 22 23 24 25 26
27 28 29 30 31