2.2.2 File Conversion:
Converting SPHERE Files to Raw Format
You must follow two critical steps to convert speech files from
a SPHERE to a
raw
format. In preparation for this step, you will need to download and
install the
NIST SPHERE software.
For this conversion process, we will use the tools w_decode and
h_strip contained in this package.
For our example, we will convert the audio file
speech.sph
from SPHERE to raw format.
- Go to the directory:
$ISIP_TUTORIAL/sections/s02/s02_02_p02/
- Convert the SPHERE file's binary data to 16-bit linear samples
using w_decode as shown in the following syntax:
 
w_decode -o pcm speech.sph speech-nb.sph
- Strip the file's header using h_strip with the following
syntax:
h_strip speech-nb.sph speech.raw
The new file is now in raw format and should be identical to
speech.raw.
Note that the only difference between these two files is that
the raw file is missing the first 1024-bytes of the SPHERE file.
As explained previously, the first 1024 bytes in a SPHERE file
contain header information.
|
|
To avoid intermediate files (e.g., speech-nb.sph in the
example above), the commands can be
piped.
You may use Unix pipes to create one command for file conversion rather
than typing three different commands. The syntax below shows the use
of the Unix pipe symbol "|" to achieve this:
w_decode -o pcm speech.sph - | h_strip - - > speech.raw
Notice that the two commands, w_decode and h_strip,
now reside in one command. The resulting file produced by each
command becomes the input for the command which follows the "|" symbol.
We typically add the downsampling process, described in
Section 2.3.1
to this pipe so that we can strip the header and downsample the data,
all in one command with no intermediate files.
|
|