Audio::SPX - Perl interface to the Sphinx audio library.

  use Audio::SPX;
  my $ad = Audio::SPX->open_sps(16000);

  $ad->start_rec or die "start_rec failed";
  $ad->stop_rec or die "stop_rec failed";
  my $samples = $ad->read($buf, $max);

  my $cad = Audio::SPX::Continuous->init($ad);
  my $cad = Audio::SPX::Continuous->init_nbfh($fh, $sps);
  my $cad = Audio::SPX::Continuous->init_raw($fh, $sps);

  $cad->calib;
  $cad->calib_loop($buf);
  $cad->set_thresh($sil, $sp);
  $cad->set_params($delta_sil, $delta_speech, $min_noise,
		   $max_noise, $winsize, $speech_onset,
		   $sil_onset, $leader, $trailer);
  my ($delta_sil, $delta_speech, $min_noise, $max_noise,
      $winsize, $speech_onset, $sil_onset, $leader, $trailer)
         = $cad->get_params;

  # If init_raw was used, this will consume the data in $buf, then
  # write back any non-slience data.  Yes, this feature is
  # undocumented in the Sphinx headers.  Yes, it's very useful.
  my $samples = $cad->read($buf, $max);

  $cad->reset;
  $cad->detach;
  $cad->attach($ad);
  $cad->read_ts;
  $cad->set_logfp(\*FH);

Warning! This interface is suboptimal and is therefore probably going to change, both in the Perl module and the underlying library.

The only supported sample rate for Audio::SPX is 16kHz (Audio::SPX::Continuous should be fine with others). init_sps() will simply fail rudely if you try something else... which means it isn't really very useful. I suggest either opening the audio device yourself, setting non-blocking mode (beware, some sound drivers don't like this...) and passing it to the init_nbfh method in Audio::SPX::Continuous, or using init_raw and managing the audio device yourself.

There isn't enough documentation yet, partly because the API is somewhat in flux, and partly because I haven't figured out what some of this stuff does either :-)

David Huggins-Daines <dhuggins@cs.cmu.edu>

perl(1), Speech::Recognizer::SPX