Jump to content

  • Log In with Google      Sign In   
  • Create Account


#ActualBacterius

Posted 28 August 2012 - 10:01 AM

If you mean generating entropy from a person speaking into some microphone by analyzing the resulting waveform, this isn't going to be trivial. The entropy source needs to be normalized so that its entropy can be measured within reasonable bounds. It is understood how to safely normalize many known entropy sources such as thermal noise, keyboard typing delays, mouse clicks, network jitter, etc... but waveforms? I have no experience with this and I suspect that it highly depends on the format...

Or do you mean something else? Could you expand on what "frequency of the sounds" mean? It isn't very clear to me.

#2Bacterius

Posted 28 August 2012 - 10:00 AM

If you mean generating entropy from a person speaking into some microphone by analyzing the resulting waveform, this isn't going to be trivial. The entropy source needs to be normalized so as to be reasonably unbiased (which is equivalent to saying there is *some* amount of entropy in what you collect). It is understood how to safely normalize many known entropy sources such as thermal noise, keyboard typing delays, mouse clicks, network jitter, etc... but waveforms? I have no experience with this and it highly depends on the format too.

Or do you mean something else? Could you expand on what "frequency of the sounds" mean? It isn't very clear to me.

#1Bacterius

Posted 28 August 2012 - 09:58 AM

If you mean generating entropy from a person speaking into some microphone by analyzing the resulting waveform, this isn't going to be trivial. The entropy source needs to be normalized so as to be unbiased (i.e. 0 occurs with the same probability as 1) and you need to take into account background noise - which, of course, is not normalized either.

Or do you mean something else? Could you expand on what "frequency of the sounds" mean? It isn't very clear to me.

PARTNERS