How To Write
Your Own
Application Programs
VOXBOX
Use
The
VOXBOX
has three routines
you can call from
BASIC:
1)
Initialization
2)
Training
3)
Recognition
They are
usually used in
this order.
Each routine
is called
by the
USR
function of Level II (or
USR0, USR1,
USR2 for
DISK
BASIC).
These
routines create the speech data files or
match
unknown
speech
data against
these
files to
find the corresponding
entry
and thus the
desired word that was spoken.
Initialization clears the
speech files, and
resets
the speech
sample
table to
all 0's.
Previously stored speech
data
is
lost.
Use
Initialization
at
the
beginning of the program once; it must
be followed
by
training.
Training accepts
known spoken words and
extracts
features from
these
words which are stored away for
future
comparison
with
unknown
words during
recognition.
During training,
a training index
number (TR=0 to
31)
is given to identify the
specific
word
sample.
Recognition
identifies unknown
words spoken to
VOXBOX.
It may
be called as
frequently as
desired
after training.
Training
samples
are
preserved during
recognition. During recognition,
speech is
presented
and
the training index number corresponding
to the
set of stored
characteristics closest to
the unknown
word is returned.
TR=32
is
reserved for "no
such
word
in
my
list."
A detailed
discussion of the use of each routine
follows
below; and a
typical program is developed step-by-step.
21
To Next Page
Loading...