 |
PM |  | |
Quote | | |
Reply | | |
Full Topic | |
| Joined: | Sep 22nd, 2008 |
| Location: | Doylestown, Pennsylvania USA |
| Posts: | 130 |
| Status: |
Offline
|
Hi Slider,
I don't know the TI chips. Are they still available? I only know of the Votrax SC-01 which was popular in the Eighties. The SpeakLet chip is controlled over a serial line; the protocol is described the manual. In the EVP-ITC group there are some members who build devices to let the SJ chip generate randomized speech. Frank Sumption uses a random voltage generator and a schmitt trigger to feed the serial input with random pulses; Rob Fitzgerald uses a programmed PIC and an EMF receiver unit to control the chip.
Stefan
Stefan,
I believe that all EVP/ITC methods are based on randomness in this dimension. The signature of noise is its true randomness. Your randomizer code is producing intelligent messages in the same manner as noise input would. The closer to random nature the more accurate may be the conversion. Your parallel printer noise input is ingenius, but he code randomizer in EVP/maker is good enough, judging from the results using the alophone file.
If a true noise source, such as your bit external generator, could be analyzed for a characteristic that might act as a squelch gate for the binary data, the nonsense alophones might blocked allowing more words to be produced in context. Phase shift, noise amplitude, bandpass, might be detected to produce a gate signal. Since my weakness is programming, I would like to try some analog tricks on my clone of the bit generator. It will keep me out of the code.
Of course, the code randomizer might be filtered by the comparison of strings of code with a library of words.
What do you think about these possibilities?
joecioppi
Last edited on Oct 13th, 2008 12:10 AM by joecioppi