The sound quality of xedencag is not particularly impressive at
the moment. All words were recorded as separate files and only after
my Mongolian colleague had left the ``studio'' (actually a table in the
kitchen) I noticed that volume and loudness showed a certain variation
across the samples, and so did the prosodic patterns. I tried to adjust
at least the volume but could do little for the rest. I also edited the
sound snippets so that the onset of speech usually starts now about 0.1
sec after the beginning of the sound sample.
The sounds were recorded using the Win95 audio recorder at a high sampling
rate (originally intended for other purposes), converted to 8kHz 8bit and
stored as *.WAV files. These were then converted to Sun audio files
(*.au) using sox. Originally I intended to use the original
22050 Hz 16bit *.WAV files but I could not convince sox to
produce meaningful output of these, and for some strange reason which I
never really probed into my Toshiba notebook's built-in sound card can
play back sounds when running Linux but fails to record them.
The conversion from *.WAV to *.au format generated a nasty
click at the onset of every sound sample to the effect that announcing
the time is accompanied by a staccato of clicks. Playing back the sound
samples under Win95's audio recorder did not reveal these clicks.