Author |
Message |
Inventor
Stream Operator
Joined: Oct 13, 2007 Posts: 6221 Location: near Austin, Tx, USA
Audio files: 267
|
Posted: Thu Jan 17, 2008 12:36 am Post subject:
Artificial Brain Waves: Hearing Machine Thought Subject description: A ChucK Neural Net with Audio-Output of it's Internal Data |
|
|
Well, as promised some time ago, here is the phoneme recognizer with the neural net outputs converted into audio form. Must use the headphones for this one to prevent audio feedback. First it asks you to pronounce A, E, I, O, and U. Then listen to the machine-like drone of audio that it makes as it trains on the data for several minutes and learns it. The drone changes over time to a somewhat more specific type of drone, but drone it is. Finally, it goes into "running..." mode where it listens to whatever you say, tries to recognize the phoneme, and makes a different neural-net sound depending on what you say.
In other words, it is the most completely useless conglomeration of excessively underutilized technology that I have ever created, lol. But I just had to find out what it sounded like when a machine thinks. Actually I'm just outputting the values of the neuron outputs using a step Unae, perhaps adding them up and clearing them each row would be more melodious.
On another note, (yes, you deserved a pun that bad today) my music website continues to get daily hits despite the fact that I deleted all external links to it and haven't mentioned it on the forum in months. I think people are looking for the boolean sequencing stuff because the Google search keywords are along those lines, some of them anyway. Just goes to show how many music enthusiasts there are out there! If you want to check out my music site, it is here:
http://www.freedomodds.com/music/
Can you name that tune in two notes? Good, because my other note is that I met with someone from the Alumni and Development office of my old University who will help me get some much needed technical writing help on my new oscillator invention. Once it is published, I'll post all about it to the proper forum of electro-music.com. Actually it turns out you can use it as a filter, a signal detector, and an amplitude modulator in addition to its primary use as a sine wave oscillator.
Well there you go, a hodge-podge of a post describing a hodge-podge of a ChucK program. Well, it was fun to do.
Inventor
Description: |
A ChucK Neural Net that learns 5 phonemes and outputs an audio representation of its internal data. Listen to the sound of a machine thinking about what you just said. |
|
Download |
Filename: |
speech2text7.ck |
Filesize: |
22.35 KB |
Downloaded: |
265 Time(s) |
|
|
Back to top
|
|
|
Dr. Spankenstein
Joined: Mar 03, 2007 Posts: 136 Location: Cambridge
Audio files: 1
|
Posted: Thu Jan 17, 2008 4:07 am Post subject:
|
|
|
Quote: |
On another note, (yes, you deserved a pun that bad today)
|
That's a couple of divisions below the "official sporks person" pun league!
I like these crazy ideas of yours for ChucK code, keep em coming.
Rhys |
|
Back to top
|
|
|
Inventor
Stream Operator
Joined: Oct 13, 2007 Posts: 6221 Location: near Austin, Tx, USA
Audio files: 267
|
Posted: Thu Jan 17, 2008 5:42 pm Post subject:
|
|
|
Thanks, Dr. Spankenstein, a little encouragement goes a long way. I wonder if I will create something tonight for you to enjoy later?... Any suggestions? |
|
Back to top
|
|
|
Dr. Spankenstein
Joined: Mar 03, 2007 Posts: 136 Location: Cambridge
Audio files: 1
|
Posted: Fri Jan 18, 2008 4:20 am Post subject:
|
|
|
How about a MIDI controlled munger/granulator using LiSa? |
|
Back to top
|
|
|
Inventor
Stream Operator
Joined: Oct 13, 2007 Posts: 6221 Location: near Austin, Tx, USA
Audio files: 267
|
Posted: Fri Jan 18, 2008 4:54 am Post subject:
|
|
|
Dr. Spankenstein wrote: | How about a MIDI controlled munger/granulator using LiSa? |
Ooh, You're getting out of my area of knowledge there, don't know MIDI and don't know LiSa. Actually I also don't know what a munger/granulator is either so I'm at a loss. I have given some thought to a sound repeater / changer thingie that takes long FFT samples, changes them somehow, and replays them in some type of purposefully erratic sequence. Is that similar to what you mean? |
|
Back to top
|
|
|
Dr. Spankenstein
Joined: Mar 03, 2007 Posts: 136 Location: Cambridge
Audio files: 1
|
Posted: Fri Jan 18, 2008 5:03 am Post subject:
|
|
|
Aye that's right, there are some examples in /examples/special/Lisa-munger
...but if that is not your cup of tea, how about a picture to sound type programme a la Sonification?
Or maybe a ambient soundscape that builds up according to what data is received through a microphone (adc)?
Rhys |
|
Back to top
|
|
|
Inventor
Stream Operator
Joined: Oct 13, 2007 Posts: 6221 Location: near Austin, Tx, USA
Audio files: 267
|
|
Back to top
|
|
|
Frostburn
Joined: Dec 12, 2007 Posts: 255 Location: Finland
Audio files: 9
|
|
Back to top
|
|
|
|