aconvert(1)							  aconvert(1)



NAME
  aconvert - convert audio data

SYNOPSIS
  aconvert [options] [inp] [out]

DESCRIPTION
  The aconvert program is used to convert between different audio file for-
  mats,	sampling rates and sample sizes.


OPTIONS
  aconvert accepts the following command line options:

  -a	  Auto Gain Control (AGC) enable.

  -b	  Byte swap input data.

  -c	  Copy raw header information from input to output when	a header is
	  skipped using	the -j option.

  -f order
	  The specified	order will be used for the FIR filter when converting
	  between different sampling rates (default is 10).

  -g gain The specified	gain will be used between input	and output files
	  (default is 1.0).

  -j bytes
	  The specified	number of bytes	will be	skipped	from the input file
	  for raw input	data (default is 0).

  -m units
	  No more than the specified number of units will be input. A value
	  of 0 implies no limit. This number is	also used to generate a
	  specific number of samples for ``fake'' inputs (sweep, tone and
	  white).

  -s	  Output statistics on standard	error.

  -v	  Enable silence compression of	input (vox). This will only work on
	  single channel input.


INP/OUT
  The inp and out arguments are	of the form:

  [name][,option...]

  Where	name may be the	desired	filename or - for stdin/stdout (default	is
  stdin/stdout if the name is left off).  option may be	any of:

  b[its]=num	  Specifies bit	width of 1 sample. Default is 8.

  c[hans]=num[:mix...]
		  Specifies input/output channels. num is the total number of
		  channels either feeding the input or feeding the output
		  (after input mixing).	Each :mix entry	specifies how the
		  input	channels are mixed into	an output channel.

		  For example: chans=4:12:34 specifies that there are 4	chan-
		  nels which will be mixed down	to 2 channels.	The first
		  will be mixed	from channels 1	and 2, while the second	will
		  be mixed from	channels 3 and 4. If you only specify the num
		  value	on input, all of the channels will be preserved	(i.e.
		  c=4 is the same as c=4:1:2:3:4).

		  If you specify more channels on output than on input,	chan-
		  nels will be duplicated. For example:	chans=2:1:2:12 will
		  take a stero stream add a third channel which	is the mix-
		  ture of streams 1 and	2. Whenever possible, down-mix/drop
		  any channels on input	(rather	than on	output)... this	is
		  MUCH more efficient.

		  If you do not	specify	the number of channels on output, all
		  channels will	be mixed into a	single output channel. The
		  default is 1 channel in and one channel out.

  e[xpon]=num	  FFT Peak enhancement factor. Default is 0.6

  [f[ormat]=]fmt  File format. Legal values are:


	  ascii	  data is read/written in ASCII.  8bit data is in hex, 16bit
		  data is in decimal and FFT vectors are in floating point.

	  htk	  Hidden Markov	Toolkit	format.

	  raw	  no header (default).

	  sphere  NIST standard	format (e.g., Timit database).


  i[incr]=samples Specifies increment between FFT windows. Default is 160.

  p[arams]=num	  Specifies number of parameters (coefficients)	per FFT	vec-
		  tor. Default is 11. This number actually includes Energy +
		  Peak-to-Peak + Zero-Crossings	+ Actual-Coeffs.

  [r[ate]=]frequency
		  Sample frequency in kilohertz. Default is 8.0.

  s[amples]=samples
		  Specifies number of samples in an FFT	window.	Default	is
		  320.

  [t[ype]=]type	  Type of data as listed below:

		  Type	     Inp/Out   Description
		  adpcm	       I/O     2,3 or [4] bits
		  alaw	       I/O     8 bit data
		  cepstrum	O      ener+ptp+zc+coeffs
		  ima	       I/O     4 bits
		  linear       I/O     16 bit data
		  melcep	O      ener+ptp+zc+coeffs
		  plp		O      ener+ptp+zc+coeffs
		  rasta		O      ener+ptp+zc+coeffs
		  sweep		I      fake linear for -m samples
		  tone		I      fake linear for -m samples
		  ulaw	       I/O     8 bit data
		  white		I      fake linear for -m samples



EXAMPLES

  Here are different ways to read from an 8 bit, 8khz, raw ULAW	file inp.snd
  and create a 16 bit, 44.1khz,	linear SPHERE format stereo audio file
  out.snd:

       aconvert	inp.snd,t=ulaw,r=8 out.au,t=linear,f=sphere,r=44.1,c=1:1:1
       aconvert	inp.snd,ulaw,8 out.au,linear,sphere,44.1,c=1:1:1
       aconvert	inp.snd	out.au,linear,sphere,44.1,c=1:1:1
       aconvert	inp.snd	-,linear,sphere,44.1,c=1:1:1 >out.au
       aconvert	- -,linear,sphere,44.1,c=1:1:1 <inp.snd	>out.au


RETURN VALUE

  none specified.

SEE ALSO

  AF(1)

BUGS
  melcep and cepstrum have not been implemented	yet.

COPYRIGHT
  Copyright 1993-1994, Digital Equipment Corporation.
  See AF(1) for	a full statement of rights and permissions.

AUTHORS
  Dave Wecker, Cambridge Research Lab, Digital Equipment Corporation.