This requires several step:
1. Set up and train the speech recog engine
2. Introduce a keyword, so that the following word is recognized as a key stroke; for instance "Key stroke F1". This avoids unwanted windows opening ( "In the last F1 race, Michael Schumacher...")
3. Using the SAPI, set up an event that triggers a callback function when "Key Stroke" is recognized.
4. Interpret the next token ("F1" in the example) using a mapping table token<->key code
5. Emulate the keystroke using an API function, for instance:
SendKeys[
^]