Techniques of Voice Disguise
T-Man
Recently while perusing a few electronic kit catalogs, I came across a voice
disguiser (Gateway Electronics, St. Louis, MO) that supposedly renders one's
voice unrecognizable. The kit uses the method of FREQUENCY INVERSION, which is
something I am not familiar with. By the looks of the circuit (pictured), it
looks like it only uses one IC and a few other caps, R's , etc. How exactly
does this frequency inversion technique work, and is there any easy/cheap way
to build one of these things. I assume that the device just changes the
spectrum of one's voice, the other features of speech such as speed, inflection
and cadence would still survive, correct? This sounds like a fun little
project to toy with (Caller ID still hasn't hit Bozeman, MT yet! :>) so any
information would be appreciated. Thanks in advance ....
__________________________________ ________________________________________
Montana State University |
Electrical Engineering | Leave Blank - For Use by Employer Only
|
Troy Curtiss (406) 994-4986 |
ieeu...@eagle.oscs.montana.edu |
__________________________________|________________________________________
Rich Taylor
----------
Michael A. Covington
>/ hpdmd48:sci.electronics / ieeu...@Eagle.oscs.montana.edu (T-Man) / 11:23 pm May 13, 1992 /
>
>Recently while perusing a few electronic kit catalogs, I came across a voice
>disguiser (Gateway Electronics, St. Louis, MO) that supposedly renders one's
>voice unrecognizable. The kit uses the method of FREQUENCY INVERSION, which is
>something I am not familiar with. By the looks of the circuit (pictured), it
Here are some ideas.
Frequency spectrum _inversion_ is a scrambling technique; it makes the voice
totally unintelligible. You then invert it again to get it back to normal.
To invert the spectrum: Mix your voice (300-3000 Hz) with a sine wave
at say 3300 Hz using a balanced modulator. Then low-pass-filter it so you
get only 300-3000 Hz at the output. Voila -- the original 300 Hz has been
mapped onto 3000 Hz, and the original 3000 onto 300, and likewise everything
in between.
To disguise your voice (to make male sound female, etc.): Invert it, then
invert it back, _but not with the same mixing frequency_. Maybe 3300 going
one way and 3600 going the other way.
A good chip to use would be the NE602 oscillator-and-balanced-mixer.
--
==========================================================================
Michael A. Covington, Ph.D. | mcov...@uga.cc.uga.edu | ham radio N4TMI
Artificial Intelligence Programs | U of Georgia | Athens, GA 30602 U.S.A.
==========================================================================
Ed Hall
>To disguise your voice (to make male sound female, etc.): Invert it, then
>invert it back, _but not with the same mixing frequency_. Maybe 3300 going
>one way and 3600 going the other way.
It might disguise your voice, but it will sound like a B-movie robot at
best (if it is intelligable at all). The harmonic relationships will be
all messed up, for reasons explained during the recent pitch-shifting
thread.
It is useful to think of the human voice as a relatively low-frequency
pulse waveform (vocal cords) followed by filters (throat, mouth, nasal
passages, etc). An ideal pulse (we're talking mathematical abstractions
here) has equal energy into all harmonics (that is, whole-number multiples
of the fundamental frequency). The output of human vocal cords is not
exactly a pulse waveform, but for frequencies of interest here, it is
close enough. If you shift all frequencies in the resulting spectrum up
or down a fixed amount, you destroy the whole-number relationship between
them. (Of course, if you shifted all frequencies by the same *fraction*,
the harmonic relationships are preserved, but as we found in the earlier
discussion, this cannot be done by simple mixing.)
There are, however, devices which can shift the pitch of a voice--and even
simulate the differences that changes in the geometry of the vocal tract
(such as between different individuals) would produce. They work by
either measuring the overall (near-) instantaneous spectrum and
manipulating the frequencies so determined (such as performing the
proportional shift mentioned above), or by assuming that the input to the
vocal tract has a relatively flat overall spectrum, determining the
filtration that, impressed upon this, would produce the actual vocal
sound, and then applying these filter parameters to a substitute input
signal. The first technique is called a "phase vocoder", while the second
is a somewhat oversimplified description of "linear predictive coding."
Both of these techniques work well for manipulating vocal material
while still preserving intelligibility. Linear predictive coding
(LPC) actually forms the basis of some highly sophisticated secure
telephone equipment, since the parameters it generates can be
sufficiently compressed to be transmitted real-time by existing modem
technology--perhaps after digital encryption.
-Ed Hall
edh...@rand.org
S. Booth
Assuming the IC is non-custom it may be simply a filter
since this is a common method of disguising a voice without
making it uninteligable.
Another possibility is that it may be somesort of frequency
multiplier/divider.
Just a couple of ideas, i may be right I may be wrong, but at least
I admit it.
--
Stephen Booth
*******************************************************************************
** Email Snail **
** ** ** bi...@uk.ac.keele.seq1 U117 **
** Barnes Hall **
** Disclamier: who would believe University of Keele **
** any of this. Not this place Keele **
** for sure. ST5 5BG **
** England **
** **
*******************************************************************************
(c) S. Booth 1990, May be transmitted by Email and USENET
only, unless text states otherwise, without permission.
But other transmission requires permision from author either in writing
or verifide Email.
Michael A. Covington
The fundamental frequency doubles, but the spectrum remains otherwise
close enough to the original that all the words are intelligible.
I've heard this described as a "chipmunk voice" effect.
Timothy J. Stanley
|
| To invert the spectrum: Mix your voice (300-3000 Hz) with a sine wave
| at say 3300 Hz using a balanced modulator. Then low-pass-filter it so you
| get only 300-3000 Hz at the output. Voila -- the original 300 Hz has been
| mapped onto 3000 Hz, and the original 3000 onto 300, and likewise everything
| in between.
|
| To disguise your voice (to make male sound female, etc.): Invert it, then
| invert it back, _but not with the same mixing frequency_. Maybe 3300 going
| one way and 3600 going the other way.
|
| A good chip to use would be the NE602 oscillator-and-balanced-mixer.
You know, in an old electric guitar special-effects book, I think
there is a circuit that does this very thing (or something mighty like
it). It is generally called a ring-modulator - but the circuit
published in the book is in fact a balanced-modulator (in spite of its
name, and I am para-quoting the author from distant memory). The name
of the book is Electronic Projects For Musicians by Craig Anderton.
You can still mail order it from the back of many guitar oriented
magazines - notably Guitar Player. You can sometimes find it in
better guitar oriented music stores for pros. A company in Indiana
(?) called PAIA also sells kits for all of the projects in the book as
well as the book, as well a a mess of other music and synthesizer
projects. I built this "ring-modulator" once from scratch (Radio
Shack has all the parts - I'd guess less than 20$ except for a 50u
cap). I never could filter out the carrier frequency (which I think
parallels the 300-3K sine wave mentioned above) completely, but I also
didn't use all the specified capacitor values because since I am an EE
I am qualified to screw up my own projects any way I want. Someone
who follows directions better than myself would probably succeed at
eliminating the carrier wave. I patched the guitar through it - it
was fun, outrageous sounding, of dubious musical merit. At extremes
of some settings, it made the guitar sound rather gong-like (really).
I never patched voice through it. As someone in one of the followups
mentioned - it might sound like a B-movie robot - that is *exactly*
what I would expect this thing to sound like with voice. In fact, as
I recall, Darth Vader in Star Wars had a bit of ring/balanced
modulator sound mixed in with James Earl Jone's real major-league
voice for that science-fiction effect. The kit I just mentioned does
have provisions to mix straight and modulated sounds.
Have fun - let us all know if you try it out.
T
Bob Kyweriga
or Radio-Electronics from the mid to late 1970's that used a one
chip design to do voice inversion.
Used a phase lock loop chip ( a 755 maybe ? ) and a handful of
passive stuff
UUCP: {crash tcnet}!orbit!pnet51!bobkay
ARPA: crash!orbit!pnet51!bob...@nosc.mil
INET: bob...@pnet51.orb.mn.org