September
1999, Issue 110
Taking
Orders:
A
Speech Recognition Module
by
John Iovine
This
versatile speech-recognition module has many uses,
ranging from video game controls to home control for
the disabled. Listen up as John explains how easy
it is to train the module and implement it in an application.
The
VoiceDirect module from Sensory is a great introduction
into the world of speech recognition. The module, shown
in Photo 1, is inexpensive and flexible.
Photo 1—This
view of the VoiceDirect module circuit board shows
the connection headers, which enable easy integration
into your next project.
|
VoiceDirect
is stand-alone capable, ready to be embedded into your
project. Or, it can be made to function as a slave under
a host processor with enhanced speech-recognition capabilities.
In
stand-alone mode, the module can recognize up to 15
words or phrases lasting up to 3.2 s each. Working as
a slave under a host CPU, the module can recognize up
to 60 words. Communication between the master CPU and
the speech-recognition module takes place over a three-wire
serial interface.
When
a trained word is recognized, the module outputs a digital
signal corresponding to the word recognized. The output
line(s) associated with the word is brought high for
1 s. This signal may be used to control external devices
with minimal external hardware.
The
module is designed to be embedded into electrical switches,
appliances, and consumer electronics. Before I delve
into the features of this particular module, let me
first define a few speech-recognition terms.
