[Octave-bug-tracker] [bug #46781] specgram flips sign of frequencies when given complex input as row vector

[anonymous]

URL:
  <http://savannah.gnu.org/bugs/?46781>

                 Summary: specgram flips sign of frequencies when given
complex input as row vector
                 Project: GNU Octave
            Submitted by: None
            Submitted on: Wed 30 Dec 2015 05:43:09 PM UTC
                Category: Octave Forge Package
                Severity: 3 - Normal
                Priority: 5 - Normal
              Item Group: Incorrect Result
                  Status: None
             Assigned to: None
         Originator Name: Alastair Harrison
        Originator Email: address@hidden
             Open/Closed: Open
         Discussion Lock: Any
                 Release: 4.0.0
        Operating System: Any

    _______________________________________________________

Details:

I don't know if Octave's specgram function is intended to allow complex input,
but it has some strange behaviour when you do.

Consider the following example. When the input is a complex column vector, we
get the expected result. When the input is a row vector, specgram actually
shows us the spectrogram of the complex conjugate of the input.


Fs = 44100;

% Generate a complex exponential wave at 8kHz
% Note that we're producing a ROW vector here
t = (0:Fs-1) / Fs;
s = exp(j*2*pi*8000*t);

% Spectrogram has no peak at 8kHz
window = ceil(40*Fs/1000);
step = ceil(20*Fs/1000);
figure(1)
specgram(s, 2^nextpow2(window), Fs, window, window-step);

% But provide the input as a COLUMN vector instead, and we get a peak
% at 8kHz
figure(2)
specgram(s.', 2^nextpow2(window), Fs, window, window-step);


The problem appears to be with this line:

  if columns(x) != 1, x = x'; endif


For complex input, the apostrophe operator both transposes and takes the
complex conjugate. A fixed version performs only a transpose:


  if columns(x) != 1, x = x.'; endif


As an aside, I believe that the matlab version of specgram detects complex
input and displays both positive and negative frequencies. It would be great
if the octave version could do something similar, rather than just chopping
off everything above the Nyquist frequency.




    _______________________________________________________

Reply to this item at:

  <http://savannah.gnu.org/bugs/?46781>

_______________________________________________
  Message sent via/by Savannah
  http://savannah.gnu.org/