pitch control

CSGnet Archive CSG1996
1

[From Bill Powers (960509.1330 MDT)]

Peter Cariani (960509.1030) --

I had thought of asking a question about pitch control, but Rick Marken
beat me to it.

In your reply, you didn't actually come up with a working model of pitch
control. Let me sketch in a generalized model, and then perhaps you
could get more specific about how the different parts would be
accomplished.

                              representation of
                                target pitch
                                      >
                                      v
                   -------> Comparison of actual -------->----
                 > pitch with target pitch |
                 > v
            Representation Representation of
                of pitch difference
               Pitch |
                ^ v
                > Conversion of pitch
            Perceptual different to action
             Function |
                ^ |
                > v
           Physical sound Measure of
           wave frequency action
                ^ |
                > >
             sound wave effect of action |
             generator <-------- on sound generator <---------

Note that this model is not dependent on how pitch is represented, or on
the method of coding of signals anywhere in the control system.

In a rate-coded system, the specific model is straightforward. Both the
perceptual signal and the reference signal are measured in impulses per
second. The reference signal is excitatory and the perceptual signal is
inhibitory; the error signal is proportional to the excess of excitation
over inhibition. For two-way action, a duplicate comparator is required
with the signs of the inputs interchanged.

From your description, an interval-coded system would be considerably

more complex. Note, however, that if the input function operated on the
basis of temporal coding patterns, it could still produce an output
perceptual signal in which the rate of firing was proportional to pitch
(although not physically having the same frequency). This would greatly
simplify the rest of the system, while still permitting the phenomena
you note to take place.

···

-----------------------------------------------------------------------
Best,

Bill P.

2

[Hans Blom, 931206]

(Gary Cziko 931203.2200 GMT)

I am soliciting help from CSGnet on a project involving manipulating
digital audio as played through the Macintosh. ... The basic goal is to
be able to play digitized speech through an interface which permits "on
the fly" control of the speed of speech without changing the pitch. ...
A graduate student, Yong Zhao, would like to use such an interface to do
a master's thesis. He is quite knowledgeable about computers, although
he has only been introduced to the Macintosh this semester. ...
If someone has some information to share on this, please let me know.

There is an almost single-chip solution to this problem. I quote from the
data sheet of the Oki MSM6322 chip:

"The MSM6322 converts in real time the pitch of the speech signal in a
range of one octave upward or downward. Two pitch control methods can be
selected. One is to change the pitch in 17 steps by two switch inputs, the
other is to select one of 16 steps by four binary input lines. Since a
microphone preamplifier and a low pass filter are built in, the pitch con-
version set can easily be configured by connecting a microphone, amplifier
and speaker in the peripheral circuit."

One of our students has used and evaluated the chip in a project to speed
up the speech output of a memo recorder for a blind typist. It works just
fine, although it has some distortion (8-bit A-D, 9 bit D-A conversion).

By the way, the chip costs about 7 dollars.

Hope this helps.

Hans