[mirtoolbox] Re: Question about fluctuation patterns
- From: Olivier Lartillot <olivier.lartillot@xxxxxxxxxxxxx>
- To: mirtoolbox@xxxxxxxxxxxxx
- Date: Thu, 6 Aug 2009 13:19:07 +0300
Hi Igor,
Igor Vatolkin kirjoitti 3.8.2009 kello 15.51:
If I calculate fluctuation patterns:
--------
res=mirfluctuation(f);
resData=mirgetdata(res);
--------
- I get a matrix of 24*1903 values. I guess 24 is the number of
bands, but I cannot understand where is the number 1903 from. It
is not divisible by 231 at any rate. What I exactly want to know
is which fluctuation pattern feature vector corresponds to which
time window (or time period) from the original song. Is it possible?
The mirfluctuation operators computes the spectrum within each of
the 24 bands separately. Each FFT here is 1903 samples long. Why?
mainly because in mirfluctuation, the mirspectrum along bands is
performed for the range 0-10 Hz and with a minimal frequency
resolution of .01 Hz.
So these 1903 samples are calculated from the complete song? Is the
number of bands constant? (E.g. for another wave I become only 23
bands..)
Yes, the 1903 samples are calculated from the complete song. The
number of bands is constant, but will still vary depending on the
original sampling rate of your audio file. I guess you obtained 23
bands for audio file with low sampling rate.
Since I want to extract FP data from different intervals of the
complete music piece (e.g. for 10 second frames), I guess I should
make something in this direction (please correct me if I'm wrong):
a = miraudio('test.wav');
f = frame(a, 'Length', 10, 's');
fluct = mirfluctuation(f);
peaks = mirpeaks(fluct,'Total',5);
However I do not get the exact amount of data which corresponds to N
peaks x M bands x K frames (5x24x30, if 5 fluctuation pattern peaks
from 24 bands for each of 30 10-second frames are calculated).
OK. Now I understand the problem. If you want a frame-decomposed
fluctuation curve, showing the temporal evolution of fluctuation frame
after frame, you can *not* use mirframe or the 'Frame' option. Why?
because mirfluctuation already implies frame decomposition from the
start, so mirframe will not toggle on the frame decomposition (already
toggled on), but just control the frame parameters.
If you want to get the temporal evolution of fluctuation, here is the
trick:
s = mirsegment('test', 0:10:100); % segmentation every 10 s. You can
put any large number as last argument (100s, or whatever).
fluct = mirfluctuation(s);
peaks = mirpeaks(fluct, 'Total', 5);
Then you get a date of the desired form N peaks x K segments x M bands.
Notice: the graphical display of the results does not work due to a
bug. I will look at the problem. Meanwhile, you can get graphical
deplay if you use the 'Summary' option in mirfluctuation, i.e.:
fluct = mirfluctuation(s, 'Summary');
Regards,
Olivier
Other related posts:
