] [Thread Prev
Re: Voice recognition
Krispen, remove the Reply To address so responses to your posts don't just
come to you. :)
I used to have an early copy of IBM's speech-to-text program, and before
moving to the UK in 2000 was having some fun getting it to make my PC obey
simple one-word commands like "shutdown", "start" and "quit". It had
a 60% chance of decoding things I said and turning them into text, and,
thinking of the audio diary that the doctor in "Lawnmower Man" used, I
dreamed of the day that I could spout on ad nauseam to make The Great Novel
or something. (Hint: it didn't happen!)
After becoming the secretary of the committee for the block of flats we
lived in, I was faced with the transcription of a number of tapes of
meetings into minutes-listings, and attempted to use the program again.
package I'd had was of course not only trained by me to understand MY
voice - but also not optimized for other than a fairly twang-free American
non-accent. Imagine what kind of a mess resulted trying to decode the
voices of a range of British folks, who have more differences between
accents in a single city than most Americans encounter in a 100-mile
The only solid end results of all this was an annoyance at what I've come
wonder is as purposeful mispronounciation, and the realization that we've
still got a long way to go before computers can understand by default an
array of accents (perhaps more than an operating system's worth of
processing for that alone!).
So I sigh, and hope for the best. For now, Dave's still not here. :)
From: "Krispen Hartung" <email@example.com>:
| Yes, I guess unless you're a fighter pilot and using cutting edge,
| government funded technology, we're stuck in the waiting game. Although,
| you'd think if I can utter "caaaaab" into my voice activated cell phone,
| with a drunken slur and a bunch of animals screaming in the background,
| and the phone successfully recognizes my voice with very minor training,
| that we could get this to work for a basic one syllable command to
| switch MIDI. :)
| Just dreaming...my foot is fine.
| -----Original Message-----
| From: Jeff Larson [mailto:Jeffrey.Larson@Sun.COM]
| Sent: Sunday, February 20, 2005 8:21 PM
| To: firstname.lastname@example.org
| Cc: Loopers-Delight@loopers-delight.com
| Subject: Re: Voice recognition
| Krispen Hartung wrote:
| > Off this topic, has anyone explored the concept of voice
| > recognition-activated MIDI switching? You could wear a head set or
| cell > phone-like mic that connected to the computer....when you said,
| let's > say, "Ring Mod", the computer would recognize the command and
| make the > MIDI change to your rack gear.
| This has been a holy grail of man/machine interfaces for a long time.
| Much progress has been made, but they still tend to be a bit quirky. I'm
| not familiar with any products specifically for musical use, but of the
| ones that I've seen, you have to have a clean quiet signal and e n u c i
| a t e v e r y c l e a r l y. Most systems require "training" where
| they make you speak a few words or phrases in order to analyze the
| characteristics of your voice. Once trained, they can only recognize a
| similar sounding voice. The difference between male and female (pitch)
| or between healthy and sick (timbre) will confuse it. When performing
| live, you have the added complication of sound bleed from other sources.
| I'm sure there are pieces you could hack together that under the right
| circumstances would work. But I'm not aware of anything that is
| packaged and optimized for use by musicians. Even if there were, it
| would still be unreliable unless you could ensure a pristine signal
| (i.e. no stage microphones).
| In practice, nothing beats a dumb old foot switch.