Data Subset of the Australian BlueFin data set:
    https://data.aad.gov.au/metadata/records/AcousticTrends_BlueFinLibrary
created by Paul Baggenstoss

200 samples of each class:
(1) BM-Ant-A 
(2) BM.Ant-B 
(3) BM.Ant-Z 
(4) Bm.D 
(5) Bp-20Hz
(6) Bp-Downsweep 

Uniform event files, Sample rate 250 Hz,   3072 samples.
Available in .wav (in directory wav), or as features
in the MATLAB files Au6ClassX.mat

Features are FFT size 384 (2/3) overlapped hanning-weighted windows,
converted to 40 linear-spaced log-energy bands.

Steps I took to create the data:

-------------------- Data selection ----------------------
For each of:
'Bm-Ant-A'    'Bm.Ant-B'    'Bm.Ant-Z'    'Bm.D'     'Bp-20Hz'    'Bp-Downsweep'

1. delete the file tags.txt
2. run the MATLAB program (e.g.) :
   $ read_au('Bm.Ant-B',256);
   This selects annotated events at random. With left mouse button
   the event is saved to 'tags.txt'
   any other mouse button, it goes to next event.

3. Since annotations are selected at random each time, there will be many duplicates.
    To get only the unique tags, run (Linux) e.g.

   $  cat tags.txt  | sort  -u > tags_BM.Ant-B.txt

   which stores unique events to 'tags_BM.Ant-B.txt' 


------------ Creating down-sampled cuts of uniform size --------------
Next, run in MATLAB for BmAntA, BmAntB, BmAntZ, BmD, Bp20, BpDwnswp

 e.g. 
  >>  scan_tags  tags_BM.Ant-B.txt    BmAntB

This creates .wav of 3072 samples long, at 250 Hz. Events of different original
sample rate are downsampled to 250 Hz.  The program also creats the file event_list.txt
which is used by csf_tk.py (The CSF Toolkit) to read in the data.
The last argument is the shortened data class name to be used later

-------------------- Extracting Features ----------------------
familiarize yourself with CSF Toolkit  http://class-specific.com/csftk
Run 
python csf_tk.py
load the data class: either  BmAntA, BmAntB, BmAntZ, BmD, Bp20, or BpDwnswp
then load the feature module 'mel3b' and compile
Then enable 'write' and calculate the features for the data classes 

--------------- Creating a binary file for PBN Toolkit ---------
familiarize yourself with PBN Toolkit  http://class-specific.com/pbntk
use MATLAB program make_binaries.m

 make_binaries('Au6Class',{'BmAntA', 'BmAntB', 'BmAntZ', 'BmD', 'Bp20', 'BpDwnswp'}, ...
  'mel3b', 'default', 1210,0,1,0,0,[24,40]);

---> Au6Class.mat

--------------- Creating the 4 data folds ---------

>> scramble_au

creates ---> Au6Classa.mat Au6Classb.mat Au6Classc.mat Au6Classd.mat

--------------- Running PBN Toolkit ---------

in PBN Toolkit, import data class, e.g.  'Au6Classa'