|
EVP with random phoneme synthesis ? A quantum leap towards direct voice ITC or just a nice idea? I have been experimenting a lot with both phoneme (speech) synthesis and random signal generators. Recently I got an "odd" idea of combining these two items to create a new type of EVP "receiver". About the phoneme synthesis: It is a rather simple program in Turbo Pascal language, and it uses PC's own speaker. I suppose that the used signal format is some kind of PWM (?). ( I'm still using the Old Good Reliable DOS in my programs!) The only disadvantage for me is that those phonemes are based on the English language and thus, when my machine speaks Finnish, so the speech sounds very "A-merry-can". The randomizer (REG) is based on a reversally coupled base/emitter junction of a normal (NPN) transistor. The outcoming signal is rather clean white noise. I've found that by strongly enhacing the RF part of that noise, gives the best results. I'm using a serial resonance circuit (f res. c.a 0,5...3 MHz) . Thus we have the Gaussian distribution very close to the theoretical one. The basic idea with random phoneme synthesis is as follows: By sampling the random binary signal to a vector of e.g. 8 bits, we may generate random numbers (binary >decimal conversion) of which only the ASCII numbers for capital letters + space are used. These ASCII numbers are used to select the appropriate phomemes for speech. The space allows a pause in the phoneme flow. We may also add some kind of modulator to the random signal path for trans-demodulation by our friends in spirit. A normal amplitude/phase modulation might be sufficient for this purpose. (?) Referring to my thoughts in an other article (see EVP_#2.doc) , so we might add a delay line (analog or digital) in the random signal path to avoid a direct observation at the modulator stage. So, here it is (or not). 7. Sept. 2001
© 2001 Martti Walden
|
||
 |
|
|