Sound filtering software.

Michael Johnson kaiser at
Wed Feb 2 10:55:59 PST 2005

On Feb 2, 2005, at 12:01 AM, Joseph Chamberlain, DDS wrote:

>Do any of you know of a software that I could use to isolate the
>of the human voice and then clean everything else to make the recording
>clearer and easier to listen to ?

<slips on audio engineer hat>

If I had a nickel for every client who came in my studio asking the
same question...

The human voice has an amazing range in frequency and in dB.  It's one
of the most versatile instruments ever.  The subtleties are amazing
when combining guttural sounds with sibilance and resonance, as well as
air flow and labial manipulation (which you probably understand being a
DDS).  This presents a great difficulty when trying to do what you've
just asked.

For example, in television and radio, you have an effective frequency
range of about 15kHz which is more than sufficient for the human voice.
 However, you're limited to 20dB of dynamic range.  This isn't a
problem if your speaker is doing a Ben Stein (Bueller?  Bueller?)
impression.  On a cassette tape, you have about 40-50dB, and on 16 bit
PCM digital audio, you have ~96dB (which is probably more than the room
you're recording in or listening in).

The human voice needs about 4kHz to be distinguishable from mumbling.
The "presence" of the human voice is (arguably) around 3.8kHz.  If you
really want to do it right, you'll need about 12kHz to get everything
without sounding like a telephone, although that could sound a bit too
sibilant if you aren't careful.  Basically, what you'll need is a
bandpass filter and an expander.  Here's what they do:

Bandpass filter - Cuts off the low end below a certain frequency and
the high end off at a different frequency.  You get for example, 150Hz
through 4kHz.  Anything below 150Hz will be dropped and anything above
4kHz will be dropped.  This should get rid of the rumble of the air
handlers down low and the rush of the air through the vents up high.
it will also sound a little muffled, but you should be able to hear
everything just fine.

Expander - Makes everything below a certain decibel level quieter.  You
set a cutoff point, for example 50dB.  Anything below that point will
unaffected.  Everything above that point will be raised according to
the ratio you set, for example 1:2, effectively making 60dB sound like
80dB (10dB above the setting at a 1:2 ratio gives a 20dB rise).

Now, if there's a lot of noise, using these simple tools is not enough.
 Many/most recording studios have a "NoNoise?" suite, or "Pro Tools?"
suite with DINR (Digidesign Intelligent Noise Reduction...pronounced
like dinner).  Each of these two applications will take a sample of the
noise, and dynamically set filtering through the entire sound file to
give you a "clean" version of the file.

The software and hardware costs thousands of dollars, but it's well
worth going to a studio where they've done this type thing before.
It's not going to clean it up like you see on the spy thriller movies,
but it can clean it up a lot to make things much more audible.

IIRC, the Nat King Cole and Natalie Cole album, "Unforgettable", was
done using Sonic Solutions NoNoise?.  The engineers took the old
recordings, cleaned up the noise, removed his vocal, and then recorded
her vocal.  When Kenny G came to Washington DC to play on BET, his
monitor engineer came to my studio because they had
lost/broken/whatever the original CD he used to "play with" Louis
Armstrong.  We loaded the song from CD, removed the saxophone at the
certain points where Kenny played the melody (using DINR), and burned a
new CD so he could play it live.

This may be a little more power than you need, but it's effectively the
same thing, and it sounds pretty good.  It also took a LOT longer to do
that that it would take to clean up your recording to make it more

Now, I'll take my audio engineer hat off and get back to being a UNIX
administrator geek...

<removes audio engineer hat and goes back to his slashdot hat>


Sometimes your joy is the source of your smile, but sometimes your
smile can be the source of your joy.

					--Thich Nhat Hanh

Do not post admin requests to the list. They will be ignored.
Scitech mailing list      (Scitech at
Help/Unsubscribe/Update your Subscription:

This email sent to eugen at

----- End forwarded message -----
Eugen* Leitl <a href="">leitl</a>
ICBM: 48.07078, 11.61144  
8B29F6BE: 099D 78BA 2FD3 B014 B08A  7779 75B0 2443 8B29 F6BE

[demime 1.01d removed an attachment of type application/pgp-signature]

More information about the cypherpunks-legacy mailing list