I asked a question in this thread a while back, but failed miserably to make the point or explanation of the question clear. I hope to try again here and make a better job explaining what I meant, or was hoping to discover:
I was imagining a device that would work IRL in some way similar to the “ignore list” on this board. If you were in company and one person was being obnoxious and loud, the device would allow you to ‘turn off’ or filter out their voice, whilst still allowing you to listen to the conversation of all the other members of the group.
Conversely, it could also be used in a loud, crowded bar to filter out all sounds other than the voice of your friends, to allow easy conversation in the midst of chaos.
(Whether this device would be a good idea or not is not for discussion in this thread.)
What I would like to find out is:
1: Are there unique characteristics of an individuals voice pattern which could be utilised to completely distinguish it from others?
2: If so, could that voice be filtered out successfully without affecting the voices of others in any way?
3: Could such a devise be built small enough to fit unobtrusively in your ears? If not, could it at least be built at all?
Thanks, yet again.
Dreams of being able to filter out the nagging missus whilst happily watching the telly…
IMNA Voice expert. However, I’ve done some voice analysis just for fun on Matlab. I don’t think voice recognition technology is mature, otherwise you’d see passwords being talked. So the voice experts are welcome to correct me.
Yes there are unique characteristics to an individual’s voice. I characterize them as objective and subjective: Objective This is the frequency pattern on a fourier transform - easily done on a computer. This can be used for voice recognition. But if you device a filter to remove this voice, you may end up removing voices of other people having similar frequencies. Subjective This is the way words are pronounced or accent of a person. Also, maybe the choice of words by individuals, the speed at which a person talks, how good the “grammar”, how the person emphasizes words, etc. etc. This can be recognized by a computer, but you’ll need to train artificial intelligence algorithms like Neural Network etc. Very difficult. 2: If so, could that voice be filtered out successfully without affecting the voices of others in any way?
IMO NO. At least not online and not with the current technology. 3: Could such a devise be built small enough to fit unobtrusively in your ears? If not, could it at least be built at all?
IMO NO again to both. Such a device may be build, but I donno if it will fit that small a size - it will need a lot of “computing power”. Also, this device will probably not work online until trained.
