Chip 1997 March/CHIP Mart 1997.iso/SesProg/ESPCH10.ZIP/MANUAL.DOC

Text File | 1995-04-14 | 9.8 KB | 293 lines

                           ECHOSPEECH


         High Quality Speech Compression for Multimedia


                        Shareware Manual

















                    Echo Speech Corporation
                    6460 Via Real
                    Carpinteria, CA  93013

                    Phone: [phone]
                    FAX:   [phone]
Copyright 1994-1995 Echo Speech Corporation.  All Rights
Reserved.

No part of this publication may be reproduced, transmitted,
transcribed, stored in a retrieval system, or translated into any
other language or computer language in whole or in part, in any
form or by any means, whether it be electronic, mechanical,
magnetic, optical, manual or otherwise, without prior written
consent of Echo Speech Corporation.

Echo Speech Corporation disclaims all warranties as to this
software, whether express or implied, including without
limitation any implied warranties of merchantability, fitness for
a particular purpose, functionality, data integrity or
protection.

ECHOSPEECH is a trademark of Echo Speech Corporation.

MS-DOS and Microsoft Windows are trademarks of Microsoft
Corporation.

See the file REGISTER.DOC for information about the benefits of
registering ECHOSPEECH.
What's ECHOSPEECH?

ECHOSPEECH is the first speech compression algorithm designed
primarily for multimedia applications.  Other speech compression
algorithms were designed for communications systems, like digital
cellular phones or military radios, so they only provide the same
narrow frequency response as a telephone, and they generally need
special hardware to run.

We at Echo Speech Corporation are in the multimedia business, so
we designed ECHOSPEECH -- efficient, high-quality, cross-platform
speech compression for multimedia.  ECHOSPEECH preserves
frequencies up to 5500 Hz, so the output sounds crisp, clear and
understandable.  ECHOSPEECH can play back speech in real time on
anything faster than a 386SX/16, with no need for a math
coprocessor.  ECHOSPEECH decompression is also available for the
Macintosh, making it the first cross-platform speech compression
package.

ECHOSPEECH reduces the amount of storage space required for 16-
bit speech by a factor of more than 13 to 1, with very little
noticeable degradation in perceived quality.  For example, for
16-bit speech sampled at 11.025 KHz, ECHOSPEECH reduces the
storage required for a second of speech from 22,050 bytes to
1,650 bytes.


Installation

This shareware version of ECHOSPEECH consists of two main
programs which run under MS-DOS.  One program is a speech
compression program called WAV2ES.EXE, which reads a Microsoft
Windows .wav file and generates a compressed speech file which
also has the file name extension .WAV.

The other program is called ES2WAV.EXE.  It reads the compressed
.WAV file and generates an uncompressed output file in .WAV
format.

Although the file name extension of the compressed speech file is
.WAV, and the file complies with Microsoft's .WAV file format, it
contains compressed ECHOSPEECH data, and must be manually
decompressed with ES2WAV.EXE before it can be played back.

This decompression can be done automatically "on the fly" with
the ECHOSPEECH Audio Compression Manager files which are part of
the registered ECHOSPEECH shareware package.

The distribution contains an install program

  INSTALL.BAT

the program files

  WAV2ES.X87
  WAV2ES.EMU
  ES2WAV.EXE
  ISFPP.EXE

and the documentation files

  LICENSE.DOC
  REGISTER.DOC
  MANUAL.DOC
  ORDER.DOC

You can place the ECHOSPEECH files in any convenient
subdirectory, or you can make a new subdirectory, like "\ECHO",
for example, and put them there.

To install ECHOSPEECH, connect to the subdirectory where you put
the ECHOSPEECH files and run the installation program,
INSTALL.BAT.  INSTALL will remind you that this version of
ECHOSPEECH is shareware, determine whether or not your system has
a math coprocessor, and copy the distribution files
appropriately.

The small program ISFPP.EXE is used by INSTALL.BAT to detect the
presence or absence of a math coprocessor.


Running ECHOSPEECH

Compressing a speech file:

To convert a Windows .WAV file containing speech sampled at
11.025 KHz to a compressed ECHOSPEECH file, type the following
command:

wav2es infile [outfile]

  wav2es - is the ECHOSPEECH coder program name.  The
  installation will have created the proper version of this
  program based on whether or not your system has a math
  coprocessor.

  infile - is the path to the input file to be compressed.  This
  file must be a Windows .WAV file which has been recorded at
  11.025 KHz.  The .WAV extension is an optional part of the file
  name on the command line and .WAV will be assumed automatically
  by the wav2es program in the absence of another extension.

  [outfile] - is the optional output file name.  If this argument
  is omitted from the command line, wav2es will use the complete
  infile path and file name, and will begin the output file name
  with the character "_".

For example, the command:

  wav2es c:\speech\sample

will cause the wav2es program to compress a file called
sample.wav which is located in the subdirectory c:\speech.  The
compressed file will be saved as c:\speech\_sample.wav.  The
wav2es program will always verify before overwriting an existing
file and give you the option of entering an alternative file
name.

Decompressing a speech file:

To convert a compressed ECHOSPEECH file to a decompressed Windows
.WAV  file, type the following command:

es2wav [/8] infile [outfile]

  es2wav - is the ECHOSPEECH decoder program.

  [/8] - This optional command line parameter tells the es2wav
  program to create an 8-bit output file (see the warning
  regarding 8-bit files below).  If you omit this parameter
  es2wav will create a 16-bit file.

  infile - is the path to the input file to be decoded.  This
  file must be a Windows .WAV file which has been previously
  coded by the wav2es program.  You can omit the .WAV extension
  as a part of the infile expression.

  [outfile] - is the optional output file name.  If this argument
  is omitted from the command line, es2wav will use the complete
  infile path, removing any initial "_" character from the file
  name, as the output file name.

For example, the command:

  es2wav /8 c:\sample\_speech

Will cause the es2wav program to decompress the file
c:\sample\_speech.wav and store the decompressed file as
c:\sample\speech.wav in 8-bit format.  If the original speech.wav
file still exists in the c:\sample subdirectory, es2wav will ask
you if it can be overwritten, and if not it will give you the
opportunity to supply an alternate file name.

A Warning About 8-Bit Speech

ECHOSPEECH will compress speech files consisting of 8-bit speech
samples, and generate files with 8-bit speech samples when it
uncompresses.  However, ECHOSPEECH and other speech compression
algorithms work much better on files of 16-bit speech samples.

If you have a 16-bit sound card, the results will be much better
if you record the speech using 16-bit samples rather than 8-bit
samples, and if you use 16-bit samples for the uncompressed
speech output from ECHOSPEECH.


A Word About ECHOSPEECH Frames

ECHOSPEECH compresses and uncompresses speech in units called
"frames."  An ECHOSPEECH "frame" is 147 samples long.  Why did we
pick 147 samples?  Because ECHOSPEECH was designed to process 75
frames of speech per second, which corresponds to 147 samples per
frame.  A rate of 75 frames per second works well with CD-ROMs
and other media.

That means that ECHOSPEECH will ignore up to 146 samples at the
end of the input raw speech file, and that output speech files
will contain a multiple of 147 samples of speech data.


Getting The Best Results With ECHOSPEECH

Remember: "garbage in, landfill out."  The better the input, the
better job ECHOSPEECH can do in retaining the quality of the
speech.

o Use good audio recording techniques.  Try to record speech with
as little background noise, hum, distortion and other extraneous
sound as possible.  Watch the gain and don't clip the speech.
Use a 16-bit sound card for recording if at all possible.

o Avoid multiple voices.  ECHOSPEECH can only handle one person
speaking at a time.

o If possible, don't use speech which has previously been
compressed and decompressed.  Like multiple generations of tape
recordings, some quality will be lost each time.

o Don't normalize the speech to full range -- this can
occasionally cause glitches in the output speech.  90% of full
range is almost always OK.

o Unlike some speech compression packages, ECHOSPEECH works just
fine on high-piched voices -- don't hesitate to use them.

o ECHOSPEECH also works well on speech which has had its dynamic
range compressed.

o ECHOSPEECH may work on some non-speech sounds, and if you like
what you get, that's great, but ECHOSPEECH is a speech coder, not
an audio coder.


Possible Problems

Even though a lot of effort has been spent perfecting ECHOSPEECH,
an occasional minor but audible glitch in the output speech is
possible  Some glitches are due to the fact that ECHOSPEECH
processes speech in units of frames.

Audible glitches can sometimes occur when the speech waveform is
changing quickly at or near a boundary between frames.  Many of
these can be eliminated simply by adding or removing a number of
zero samples (say 50 to 75) at the beginning of the speech file.

Other glitches may result when the input speech has been clipped
or otherwise distorted.

Comments about ECHOSPEECH can be sent to [e-mail]