Converting matlab code to python

Question

Okay so I have a datafile from an EEG scan (a binary file, data.eeg), and in matlab the code to read the file and plot a section of the data goes like this:

sr=400;                                                     % Sample Rate
Nyq_freq=sr/2;                                              % Nyquist Frequency
fneeg=input('Filename  (with path and extension) :', 's');  
t=input('How many seconds in total of EEG ? : ');
ch=input('How many channels of EEG ? : ');
le=t*sr;                                                    % Length of the Recording
fid=fopen(fneeg, 'r', 'l');                                 % Open the file to read 
EEG=fread(fid,[ch,le],'int16');                             % Read Data -> EEG Matrix
fclose ('all');     
plot(EEG(:,3))

Here is my attempt to "translate"

from numpy import *
from matplotlib.pylab import *

sample_rate = 400
Nyquist = sample_rate/2.
fneeg = raw_input("Filename (full path & extension): ")
t = int(raw_input("How many secs in total of EEG?: "))
ch = int(raw_input("How many channels of EEG?: "))
le = t*sample_rate
fid = open(fneeg, 'r')
EEG = fromfile(fneeg, int16)

Here's where things get confusing to me. According to the documentation, matlab's fread is a method of reading binary files via fread(loaded_file, size, data_type). The alternative in python is using numpy's fromfile and reshaping (according to this thread here: MATLAB to Python fread) using the built in reshape function. I'm not sure how this works, or even relates to the matlab method? I'm sorry if my question is confusing, matlab is still very new to me

Edit: If you wanna look have a look at the file here it is: https://www.dropbox.com/s/zzm6uvjfm9gpamk/data.eeg

Edit2: The answers to the raw inputs are t=10, ch=32. In fact I'm not sure why I'm even asking for user input now that I think about it..

It sounds like you've already answered your own question. As near as I can tell, you just want eeg = numpy.fromfile(filename, 'int16').reshape(ch, le). Also, you can skip explicitly opening the file, if you'd like. fromfile will accept a filename as well as a file object. — Joe Kington
– Joe Kington, Commented Nov 12, 2013 at 19:38
I've added a link to the data in the edit. @JoeKington, that was my reasoning too, but why is it that we need to reshape? I don't understand the reasoning, you know what I mean? — faskiat
– faskiat, Commented Nov 12, 2013 at 19:55
@Norman, np.fromfile just returns a 1d array (of length ch*le, but you want it to be a ch X le matrix. reshape simply makes this change, without touching the data itself. — askewchan
– askewchan, Commented Nov 12, 2013 at 20:39
@NormanB, do you happen to be using Neuroscan? I'm trying to import Neuroscan files (.cnt or .eeg) to python and running into a lot of issues. — Amyunimus
– Amyunimus, Commented Jun 13, 2014 at 22:40

Amyunimus · Accepted Answer · 2014-06-13 22:41:57Z

2

As discussed in the comments by yourself and @JoeKington, this should work (I removed the input stuff for testing)

import numpy as np

sample_rate = 400
Nyquist = sample_rate/2.0
fneeg = 'data.eeg'
t = 10 
ch = 32
le = t*sample_rate
EEG = np.fromfile(fneeg, 'int16').reshape(ch, le, order='F')

Without the reshape, you get:

In [45]: EEG
Out[45]: array([ -39,  -25,  -22, ..., -168, -586,  -46], dtype=int16)

In [46]: EEG.shape
Out[46]: (128000,)

With reshaping:

In [47]: EEG.reshape(ch, le, order='F')
Out[47]: 
array([[ -39,  -37,  -12, ...,    5,   19,   21],
       [ -25,  -20,    7, ...,   20,   36,   36],
       [ -22,  -20,    0, ...,   18,   34,   36],
       ..., 
       [ 104,  164,   44, ...,   60,  -67, -168],
       [ 531,  582,   88, ...,   29, -420, -586],
       [ -60,  -63,  -92, ...,  -17,  -44,  -46]], dtype=int16)

In [48]: EEG.reshape(ch, le, order='F').shape
Out[48]: (32, 4000)

edited Jun 13, 2014 at 22:41

Amyunimus

1,6334 gold badges21 silver badges44 bronze badges

answered Nov 12, 2013 at 20:43

askewchan

46.7k18 gold badges125 silver badges135 bronze badges

Sign up to request clarification or add additional context in comments.

7 Comments

askewchan Over a year ago

Haha, you're welcome, but the credit is due to you and @Joe :P

Joe Kington Over a year ago

NormanB - Just on a side note, it's possible (? I can't remember.) that matlab uses column-major ordering instead of row major... en.wikipedia.org/wiki/Row-major_order In that case, the results were probably written to disk in column-major order, and you'd do fromfile(...).reshape(le, ch).T (note that le and ch are swapped). Just something to be aware of, at any rate. If your data looks garbled, give it a try.

faskiat Over a year ago

Oh yeah you're right it does(en.wikipedia.org/wiki/Row-major_order). Good catch!

askewchan Over a year ago

That's actually copied and pasted from ipython (not notebook). A nice compromise between ipython notebook and the standard python interpreter.

askewchan Over a year ago

Nice catch, @Joe! An easier fix is to use order='F'. I've edited to reflect it.

|

Collectives™ on Stack Overflow

Converting matlab code to python

1 Answer 1

7 Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

7 Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related