| Title: | Atari ST, TT, & Falcon |
| Notice: | Please read note 1.0 and its replies before posting! |
| Moderator: | FUNYET::ANDERSON |
| Created: | Mon Apr 04 1988 |
| Last Modified: | Tue May 06 1997 |
| Last Successful Update: | Fri Jun 06 1997 |
| Number of topics: | 1433 |
| Total number of notes: | 10312 |
Hello there,
Does anyone have any info on voice-recognition and/or speech
synthesis hardware or software for the atari st.
I've got an Atari ST 520STFM and I've developed and natural language
parser system which does some query translation for database
systems.
It would be nice to just have to speak the request instead of
having to type it in.
Any help appreciated,
John Browne
| T.R | Title | User | Personal Name | Date | Lines |
|---|---|---|---|---|---|
| 482.1 | I'm doing some experimentation now | BIGELO::PETERS | Don Peters, CTS1-2/H6, 287-3742 | Wed May 17 1989 15:18 | 17 |
Interesting you should ask, since I am doing some experimentation in that area now. Essentially, my first step is to take digitized input and generate a spectrogram, or a plot/graph/image that shows time on the horizontal axis and frequency on the vertical axis. I don't think FFT routines are useful since I'm not handling repetitive waves. Right now, I'm examining the spectrums of square pulses, triangular pulses, sine wave wave trains, and swept sine waves. I'm learning about frequency resolution verses time resolution tradeoffs. All my experimentation is done in GFA Basic V3, since the modification/test cycle is very fast. I also recently bought a Digisound device that plugs into the cartridge port and digitizes sound, storing it in memory. This speech data is really what I want to analyze so that I can apply some pattern recognition software to it. But, I realize many, many, have gone before me with not much to show for it. At least I'm having fun learning some important principles. | |||||
| 482.2 | Useful in any case | BENTLY::MESSENGER | Badness comes in waves. | Thu May 18 1989 17:10 | 10 |
Re: .-1
> I don't think FFT routines are useful
> since I'm not handling repetitive waves.
I believe FFT will convert from time -> frequency domain regardless of
what the signal is. You can analyze *noise* with a Fourier
transformation.
- hbm
| |||||