How do I play a sound in Octave?

问题:

Octave appears to assume that a specific sound playing utility will be available on a system but doesn't seem to provide the ability to specify an alternate. In the error below, Octave is looking for ofsndplay, which is not a utility available on all systems.

octave:38> sound(beamformed_20)

sh: ofsndplay: command not found

Is there an Octave configuration setting or code fragment that I can use to specify an system-appropriate utility?

回答1:

I've overridden the playaudio function from octave with the following function. This will work only after installing sox.

sudo apt-get install sox

(in ubuntu)

function [ ] = playaudio (x, sampling_rate)

    if nargin == 1
        sampling_rate = 8000
    end
    file = tmpnam ();
    file= [file, '.wav'];
    wavwrite(x, sampling_rate, file);
    ['play ' file ]
    system(['play ' file ]);
    system(['rm ' file]);
end

A similar approach will allow you to record too:

% Copyright (C) 1995, 1996, 1997, 1998, 1999, 2000, 2003, 2004, 2005,
%               2006, 2007 John W. Eaton
%
% This file is part of Octave.
%
% Octave is free software; you can redistribute it and/or modify it
% under the terms of the GNU General Public License as published by
% the Free Software Foundation; either version 3 of the License, or (at
% your option) any later version.
%
% Octave is distributed in the hope that it will be useful, but
% WITHOUT ANY WARRANTY; without even the implied warranty of
% MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
% General Public License for more details.
%
% You should have received a copy of the GNU General Public License
% along with Octave; see the file COPYING.  If not, see
% <http://www.gnu.org/licenses/>.

% -*- texinfo -*-
% @deftypefn {Function File} {} record (@var{sec}, @var{sampling_rate})
% Records @var{sec} seconds of audio input into the vector @var{x}.  The
% default value for @var{sampling_rate} is 8000 samples per second, or
% 8kHz.  The program waits until the user types @key{RET} and then
% immediately starts to record.
% @seealso{lin2mu, mu2lin, loadaudio, saveaudio, playaudio, setaudio}
% @end deftypefn

% Author: AW <[e-mail]>
% Created: 19 September 1994
% Adapted-By: jwe
% And adapted again 11/25/2010 by Rob Frohne    
function X = record (sec, sampling_rate)


  if (nargin == 1)
    sampling_rate = 8000;
  elseif (nargin != 2)
    print_usage ();
  endif

  file = tmpnam ();
  file= [file,".wav"];

  input ("Please hit ENTER and speak afterwards!\n", 1);

  cmd = sprintf ("rec -c1 -r%d %s trim 0 %d",
                   sampling_rate, file, sec)

  system (cmd);

  X = wavread(file);

end

回答2:

Install alsa-utils or pulseaudio-utils and put the following in your ~/.octaverc:

global sound_play_utility = 'aplay';

or

global sound_play_utility = 'paplay';

回答3:

On one of my Linux machines, I created the following ofsndplay script to work around the hard-wired dependency:

$ cat /usr/bin/ofsndplay

#!/bin/sh
## Coping with stupid dependency on ofsndplay in octave
play -t au -

This particular script uses the SoX play utility.

Admittedly, the comment is unnecessary for the functionality but it certainly made me feel better....

回答4:

playaudio is broken!

It's worth reading the default implementation of playsound (version 3.6.2):

function playaudio (name, ext)

  if (nargin < 1 || nargin > 2)
    print_usage ();
  endif

  if (nargin == 1 && isnumeric (name))
    ## play a vector
    if (! isvector (name))
      error ("playaudio: X must be a vector");
    endif
    X = name(:) + 127;
    unwind_protect
      file = tmpnam ();
      fid = fopen (file, "wb");
      fwrite (fid, X, "uchar");
      fclose (fid);
      [status, out] = system (sprintf ('cat "%s" > /dev/dsp', file));
      if (status != 0)
        system (sprintf ("paplay --raw \"%s\"", file))
      endif
    unwind_protect_cleanup
      unlink (file);
    end_unwind_protect
  elseif (nargin >= 1 && ischar (name))
    ## play a file
    if (nargin == 1)
      name = [name ".lin"];
    elseif (nargin == 2)
      name = [name "." ext];
    endif
    if (any (strcmp (ext, {"lin", "raw"})))
      [status, out] = system (sprintf ('cat "%s" > /dev/dsp', name));
      if (status != 0)
        system (sprintf ('paplay --raw "%s"', name))
      endif
    elseif (any (strcmp (ext, {"mu", "au" "snd", "ul"})))
      [status, out] = system (sprintf ('cat "%s" > /dev/audio', name));
      if (status != 0)
        system (sprintf ('paplay "%s"', name))
      endif
    else
      error ("playaudio: unsupported extension '%s'", ext);
    endif
  else
    print_usage ();
  endif

endfunction

There are some things to note:

1. Writing directly to /dev/dsp always fails in the latest linux distros, so every time you run the command you will get an error (on the cat > /dev/dsp line).
2. It's hard-coded to use paplay, a command-line pulseaudio player.
3. The paplay call will never work because paplay defaults to s16ne (probably a typo; I think they meant s16be - signed 16-bit big endian), and playaudio writes unsigned 8-bit!
4. It calls it using system(). Always a bad idea.
5. It writes the audio out to a file rather than streaming it. Might cause problems for large files.
6. Unlike matlab, it doesn't handle floating point audio. It actually only supports 8-bit audio! Kind of stupid since that is the result returned by wavread!
7. Unlike matlab, it only supports one sample rate (44100 Hz).

This function is extremely hacky, insecure and unreliable. If it in any way represents the code quality elsewhere in Octave... well, that is worrying. It should really really be reimplemented as proper function in Octave using portaudio.

Slightly better version

I don't really have time or motivation to do a lot of hacking on octave, so in the mean time I suggest you instead use this slightly better function:

function playsound(wav, samplerate)
  # Play a single-channel wave at a certain sample rate (defaults to 44100 Hz).
  # Input can be integer, in which case it is assumed to be signed 16-bit, or
  # float, in which case it is in the range -1:1.

  if (nargin < 1 || nargin > 2)
    print_usage();
  endif

  if (nargin < 2)
    samplerate = 44100;
  end

  if (!isvector(wav))
    error("playsound: X must be a vector");
  endif

  # Write it as a 16-bit signed, little endian (though the amaaazing docs don't say the endianness)

  # If it is integers we assume it is 16 bit signed. Otherwise we assume in the range -1:1
  if (isfloat(wav))
    X = min(max(wav(:), -1), 1) * 32767; # Why matlab & octave do not have a clip() function... I do not know.
  else
    X = min(max(wav(:), -32767), 32767) + 32767;
  endif
  unwind_protect
    file = tmpnam ();
    fid = fopen (file, "wb");
    fwrite (fid, X, "int16");
    fclose (fid);
    # Making aplay (alsa) the default, because let's be honest: it is still way more reliable than
    # the mess that is pulseaudio.
    if (exist("/usr/bin/aplay") == 2)
      system(sprintf("/usr/bin/aplay --format=S16_LE --channels=1 --rate=%d \"%s\"", samplerate, file))
    elseif (exist("/usr/bin/paplay") == 2)
      system(sprintf("/usr/bin/paplay --format=s16le --channels=1 --rate=%d --raw \"%s\"", samplerate, file))
    endif
  unwind_protect_cleanup
    unlink (file);
  end_unwind_protect

endfunction

This is still a very hacky function. But it should be at least a little bit more reliable than playaudio! I'll leave an implementation of soundsc as an exercise for the reader.

回答5:

I am on a Mac (Yosemite), and discovered a simpler solution than what others have suggested. Just in case this is still relevant for anybody:

First install SoX: http://sox.sourceforge.net/

(via Homebrew)

brew install sox

Now on the terminal command line you can use:

play “/path/to/sound file.wav"

...and you will hear beautiful music.

But that command does not work from within Octave. This does work:

system(‘play “/path/to/sound file.wav”’);

回答6:

On OSX, this is what I did to get sound working:

from the sound command help:

This function writes the audio data through a pipe to the program "play" from the sox distribution. sox runs pretty much anywhere, but it only has audio drivers for OSS (primarily linux and freebsd) and SunOS. In case your local machine is not one of these, write a shell script such as ~/bin/octaveplay, substituting AUDIO_UTILITY with whatever audio utility you happen to have on your system: #!/bin/sh cat > ~/.octave_play.au SYSTEM_AUDIO_UTILITY ~/.octave_play.au rm -f ~/.octave_play.au and set the global variable (e.g., in .octaverc) global sound_play_utility="~/bin/octaveplay";

I named the following script "octaveplay" and put it in ~/bin:

cat > ~/.octave_play.aif
afplay ~/.octave_play.aif
rm -f ~/.octave_play.aif

Then I created .octaverc and added: global sound_play_utility="~/bin/octaveplay";

Voila!

回答7:

On Octave 4.2.1. You can play a wav file as follows

Save the following code in a file playWav.m

function playWav(inputFilePath)
  [y, fs] = audioread(inputFilePath);
  player  = audioplayer(y, fs);
  playblocking(player)
end

Then you can call the function as playWav('/path/to/wavfile'); from Octave commandline.

Tested on Windows 7.