« Up To His Eyeballs In It | Main | Ayers Was Close Friends With an Islamist Who Was Close Friends With... »

October 10, 2008


Twenty years ago I worked on a cockpit speech recognition system. Speech recognition at that time had to cope with some complications. Airflow around the canopy creates a lot of noise, but this noise is attenuated by the air mask wherein the microphone was located. However, the microphone (back then) was located way too close to the regulator. This caused a worse regulator noise problem. Finally, the presence of the air mask on the pilot's face changes the spectrum of his voice. He has to talk against back-pressure. And his voice hits the inside of the air mask and is changed by that, too.

Unless the tech has gotten a lot better, cockpit voice systems will use a constrained vocabulary and grammar. The pilot's discourse will be restricted to the business of flying the plane, managing displays, and setting up modes of attack. Also, pilots should be able to train the system to respond to the peculiarities of their own voices.

The coolest thing I ever saw was the work on "evoked potentials" that was Firefox stuff (not the browser, the Clint Eastwood movie). I don't know if it ever came to anything. The movie was fun, but the bit about thinking in Russian was exceptionally bad bad science-wise.

The comments to this entry are closed.


  • The 2006 Weblog Awards
  • "This is a great blog."



  • Before posting a comment, ask yourself whether it is polite, fair, and truthful. Comments are auto-deleted if they contain profanity (even with ast*ri*ks). Comments may also be edited or deleted if they include anything false, misleading, insulting, unethical, illogical or spamlike. Rude comments or spam result in a permanent ban of future comments.