SAPI Client
Interface
Brainhat has a SAPI,
version 4 speech recognition/ speech generation client. You will
find it as bhatclnt.exe in the distribution directory.
The client lets you talk to the computer, and the
lets the computer talk back to you, using Brainhat as the
intelligent agent.
- You need a speech engine: If
you have a SAPI compliant speech engine, you can download the
client and give it a try. See sapi
for the SAPI SDK. Many speech engines, including Dragon
Systems and ViaVoice can interface via SAPI.
- You may also need the Winsock2
update from Microsoft if you are running Windows95.
SAPI client traffic takes place
over port 4144. This is the port that compuserve uses. I chose to
make use of it too because it is one of the ports that is opened
through a number of corporate firewalls.
The client makes use of a dictation
grammar, supporting almost any utterance the speaker might
voice. This makes the interface more flexible than one supporting
a limited or context free grammar. It also makes it much
more susceptible to mis-interpretation of something the user may
have said. To try to increase the recognition rate, Brainhat
evaluates the first handful of possible interpretations returned
by the engine, and it pumps hints to the engine to try to focus
the interpretation. However, there is much more to be done. The
better your speech engine recognizes your voice, the better.
The mp3 demo on the opening page
was created using this client.
|