| Page 2 of 2 | 1 | 2 |
|
|
"Bill Janssen" wrote in message ... Steve Nosko wrote: Hmmmm... Isn't this impossible since the spectrum has some average over time thus corrupting the time element? I was wondering if this would be possible also. But I suppose that if the sample sets overlapped and you did a lot of sets you could convert back to the original signal. Would require a lot of computing though. interesting thought. Bill K7NOM Seems to me you'll need spectrum samples at often enough time intervals to have enough of the original voice formant samples to re-create the time domain form...but this is off the cuff. Been years since studying Fourer and speesh recognition and synthesis. Actually doing it...now that's another story. Speculation is the easy part... 73, -- Steve N, K,9;d, c. i My email has no u's. |
"Bill Janssen" wrote in message ... Steve Nosko wrote: Hmmmm... Isn't this impossible since the spectrum has some average over time thus corrupting the time element? I was wondering if this would be possible also. But I suppose that if the sample sets overlapped and you did a lot of sets you could convert back to the original signal. Would require a lot of computing though. interesting thought. Bill K7NOM Seems to me you'll need spectrum samples at often enough time intervals to have enough of the original voice formant samples to re-create the time domain form...but this is off the cuff. Been years since studying Fourer and speesh recognition and synthesis. Actually doing it...now that's another story. Speculation is the easy part... 73, -- Steve N, K,9;d, c. i My email has no u's. |
"Bill Janssen" wrote in message ... Steve Nosko wrote: Hmmmm... Isn't this impossible since the spectrum has some average over time thus corrupting the time element? I was wondering if this would be possible also. But I suppose that if the sample sets overlapped and you did a lot of sets you could convert back to the original signal. Would require a lot of computing though. interesting thought. Bill K7NOM Seems to me you'll need spectrum samples at often enough time intervals to have enough of the original voice formant samples to re-create the time domain form...but this is off the cuff. Been years since studying Fourer and speesh recognition and synthesis. Actually doing it...now that's another story. Speculation is the easy part... 73, -- Steve N, K,9;d, c. i My email has no u's. |
Not sure if this suggestion has come up yet, but here goes.
The visual spectrum goes from low frequencies in the red continuously through to violet at the other end, with I believe a frequency change of 2x, meaning the violet colors have twice the frequency of the red ones. Why not take that spectrum and directly convert it to a sound by running it through a frequency domain to time domain transformation; I believe that's called an inverse Fourier transform? It would be exactly the same as taking the frequency spectrum of a sound and getting the actual sound back by running it through the transform function. I think this is a neat project. I'd be willing to take it on, with a paycheck in return of course. ^_^ -- Sincerely, --- Dave ---------------------------------------------------------------------- It don't mean a thing unless it has that certain "je ne sais quoi" Duke Ellington ---------------------------------------------------------------------- "Steve Nosko" wrote in message ... "Bill Janssen" wrote in message ... Steve Nosko wrote: Hmmmm... Isn't this impossible since the spectrum has some average over time thus corrupting the time element? I was wondering if this would be possible also. But I suppose that if the sample sets overlapped and you did a lot of sets you could convert back to the original signal. Would require a lot of computing though. interesting thought. Bill K7NOM Seems to me you'll need spectrum samples at often enough time intervals to have enough of the original voice formant samples to re-create the time domain form...but this is off the cuff. Been years since studying Fourer and speesh recognition and synthesis. Actually doing it...now that's another story. Speculation is the easy part... 73, -- Steve N, K,9;d, c. i My email has no u's. |
Not sure if this suggestion has come up yet, but here goes.
The visual spectrum goes from low frequencies in the red continuously through to violet at the other end, with I believe a frequency change of 2x, meaning the violet colors have twice the frequency of the red ones. Why not take that spectrum and directly convert it to a sound by running it through a frequency domain to time domain transformation; I believe that's called an inverse Fourier transform? It would be exactly the same as taking the frequency spectrum of a sound and getting the actual sound back by running it through the transform function. I think this is a neat project. I'd be willing to take it on, with a paycheck in return of course. ^_^ -- Sincerely, --- Dave ---------------------------------------------------------------------- It don't mean a thing unless it has that certain "je ne sais quoi" Duke Ellington ---------------------------------------------------------------------- "Steve Nosko" wrote in message ... "Bill Janssen" wrote in message ... Steve Nosko wrote: Hmmmm... Isn't this impossible since the spectrum has some average over time thus corrupting the time element? I was wondering if this would be possible also. But I suppose that if the sample sets overlapped and you did a lot of sets you could convert back to the original signal. Would require a lot of computing though. interesting thought. Bill K7NOM Seems to me you'll need spectrum samples at often enough time intervals to have enough of the original voice formant samples to re-create the time domain form...but this is off the cuff. Been years since studying Fourer and speesh recognition and synthesis. Actually doing it...now that's another story. Speculation is the easy part... 73, -- Steve N, K,9;d, c. i My email has no u's. |
Not sure if this suggestion has come up yet, but here goes.
The visual spectrum goes from low frequencies in the red continuously through to violet at the other end, with I believe a frequency change of 2x, meaning the violet colors have twice the frequency of the red ones. Why not take that spectrum and directly convert it to a sound by running it through a frequency domain to time domain transformation; I believe that's called an inverse Fourier transform? It would be exactly the same as taking the frequency spectrum of a sound and getting the actual sound back by running it through the transform function. I think this is a neat project. I'd be willing to take it on, with a paycheck in return of course. ^_^ -- Sincerely, --- Dave ---------------------------------------------------------------------- It don't mean a thing unless it has that certain "je ne sais quoi" Duke Ellington ---------------------------------------------------------------------- "Steve Nosko" wrote in message ... "Bill Janssen" wrote in message ... Steve Nosko wrote: Hmmmm... Isn't this impossible since the spectrum has some average over time thus corrupting the time element? I was wondering if this would be possible also. But I suppose that if the sample sets overlapped and you did a lot of sets you could convert back to the original signal. Would require a lot of computing though. interesting thought. Bill K7NOM Seems to me you'll need spectrum samples at often enough time intervals to have enough of the original voice formant samples to re-create the time domain form...but this is off the cuff. Been years since studying Fourer and speesh recognition and synthesis. Actually doing it...now that's another story. Speculation is the easy part... 73, -- Steve N, K,9;d, c. i My email has no u's. |
Steve Nosko says... ...and speesh recognition... Had a bit too much to drink before posting? :) Sorry. Couldn't resist... |
Steve Nosko says... ...and speesh recognition... Had a bit too much to drink before posting? :) Sorry. Couldn't resist... |
Steve Nosko says... ...and speesh recognition... Had a bit too much to drink before posting? :) Sorry. Couldn't resist... |
| All times are GMT +1. The time now is 03:12 PM. |
|
Powered by vBulletin® Copyright ©2000 - 2025, Jelsoft Enterprises Ltd.
RadioBanter.com