Hello there
I was thinking to try to reuse the MPTK (Matching Pursuit ToolKit) libraries available on the IRISA website:
http://mptk.irisa.fr/
... to make a matching pursuit algorithm box for multichannel time-frequency analysis.
Building these MPTK source produces a mptk.dll library that may ease the work.
But looking to the outputs of the algorithm, basically a distribution of atoms (the basic waveforms of the selected dictionnary), I wondered how to display this distribution of in the O.V. designer. The MPTK libraries include an executable called MPviewer whose task is to build a time-frequency map from this distribution of the atoms retained to describe the EEG signals flow. For each chunck of data (or if small, several ones concatenated into longer signals) processed with MPTK, you would get a time-frequency block.
I don't know if I am correct but O.V. seems to build the time-frequency display by stacking the spectra (a vector of amplitudes / phases for each channel) of the received chuncks of data ? Here, we would have to stack time-frequency blocks, not a vector for each channel but a matrix for each channel. Is it possible directly with the available O.V. tools (or by tweaking them slightly) ?
Jean-François B.
Implementing a matching pursuit in OpenVibe
Re: Implementing a matching pursuit in OpenVibe
Hello Jeff_B,
There is no visualization capable of displaying several matrices in the way you describe in OpenViBE (yet). However it should be fairly easy to modify the time-frequency display to do what you want, provided you have some experience with Gtk.
Cheers
Jozef
There is no visualization capable of displaying several matrices in the way you describe in OpenViBE (yet). However it should be fairly easy to modify the time-frequency display to do what you want, provided you have some experience with Gtk.
Cheers
Jozef