Forum: help

RE: mprecons [ Reply ]
By: Rémi Gribonval on 2008-09-17 20:23

>Why for 1000 atoms is there only 2 different length of atoms?

The dictionary used to perform the decomposition contains only atoms at two scales, as described in the dictionary description file dic_gabor_two_scales.xml

>the windowtype is 9, what does that mean?
This is an integer which defines the shape of the window. 9 seems to stand for Gaussian, the shape specifie in dic_gabor_two_scales.xml

RE: mprecons [ Reply ] By: chris layman on 2008-09-17 19:32	[forum:98491]
just bumping this up, any help on this perhaps?

RE: mprecons [ Reply ]
By: chris layman on 2008-09-09 19:23

[forum:98445]

ok...great that work.

Next question. I am playing around with the glockenspiel.wav

the parameter 'len' is the width of gabor atom, correct? My question is when decomposing this signal say for 1000 atoms, the book is partitioned into two sets. One set is for atoms all of one length (len=128) and another set (len = 16384). Why for 1000 atoms is there only 2 different length of atoms?

also the windowtype is 9, what does that mean?

thanks
chris

RE: mprecons [ Reply ] By: Rémi Gribonval on 2008-09-09 11:44	[forum:98438]
mprecons(book,residual) reconstructs the signal from the book plus the residual. It is good that it matches the original signal. To reconstruct just the signal from the book, use mprecons(book). Cheers, Remi.

mprecons [ Reply ]
By: chris layman on 2008-09-09 02:38

[forum:98433]

Hi

Any advice about this error.

I run the test signal from the GettingStarted ;
[sig sampleRate] = sigread(mptkInfo.path.exampleSignal);
dict = dictread(mptkInfo.path.defaultDict);
[book residual decay] = mpdecomp(sig,sampleRate,dict,100);
figure(5);bookover(book,sig);

then i want to reconstruct the signal with the residuals with,
>> [sg]=mprecons(book,residual);

sg should be the reconstructed signal with 100 atoms ?
but when i plot it against the original signal sig, they are the same.

thanks

-chris