CONTRACT RD-94 TASK ORDER NO. 2

• Document Type: CREST
• Collection: General CIA Records
• Document Number (FOIA) /ESDN (CREST): CIA-RDP78-03300A001900110027-9
• Release Decision: RIPPUB
• Original Classification: S
• Document Page Count: 3
• Document Creation Date: December 22, 2016
• Document Release Date: May 3, 2011
• Sequence Number: 27
• Publication Date: September 18, 1956
• Content Type: LETTER

Sanitized Copy Approved for Release 2011/05/03: CIA-RDP78-0330OA001900110027-9 ,
UUI ITIUU L L 25X1
8sp t.*bor 18, 1956
25X
25X
sv3MsarI
Coatrsot RD?
task terdsr ft o 2
In s toads. with Article 2 of the basis sontnaet,
there, are for &vdod borewith two (2) eoptes of the lroat
a Report top August* 1954 on ask Ordw goo a 00,t%944,
~94.
liN+e
opw t is dstod $.pteetiber , 1956. This rs Ort is -ci1 *al_
nW. La s;"itisasl OW to bsiaa bold is b dare 25X1
p Jest .a L*.e for the use of yeas' porsom:ol st-ils at this
3e ~tisa~.
In ss ,stioa srith this aomtbly progress rrport,
Who -iag tats ration Is ssbwibtsdz
T al Mrss to 74031-56 $ 2,3U
ososttassts as of
- Ohl" as of 7~-3-
l4~rts
Tsry tray yam,
trasttstreliea
OM
msva
%A
~~'~? ..,........ JNPT-~;QNFIDENTIAL
Sanitized Copy Approved for Release 2011/05/03: CIA-RDP78-0330OA001900110027-9
 Sanitized Copy Approved for Release 2011/05/03: CIA-RDP78-0330OA001900110027-9
MONTHLY PROGRESS REPORT
August 1956
Task Order No. 2
Contract No. RD-94
AUDIO NOISE REDUCTION CIRCUITS
The object of this project is to develop a noise reduction
circuit suitable for use in separating speech intelligence from
a signal containing speech and noise when the speech intelligence
is masked by the noise. The proposed method consists of passing
the signal through several frequency selective channels. Each
channel contains a non-linear element which discriminates against
the noise when the speech information in the channel is greater
than the noise.
During August the construction of a four channel noise re-
duction circuit has been completed. The circuit is arranged so
that the signal in each channel may be passed through directly, or
it may be passed through a non-linear element whose threshold is
adjustable. Circuit adjustments have been completed so that the
four channel system performs satisfactorily.
Using the four channels, without the non-linear element feature,
experiments have been performed aimed at ascertaining the most
desirable bandwidths and band spacings required to give adequate
noise reduction.
First, listening tests were made with various speech samples
mixed with white noise. The samples were passed through several
different combinations of bandpass filters. In each case studied
it was found that a wide band system gave the greatest intelligi-
bility for a given signal to white noise ratio.
Second, listening tests were performed to determine the
masking effect of noise not in the speech pass band. In this ex-
periment the speech was contained in a noise free band between
700 and 3500 cps. Noise was then introduced outside this band.
First, through a low pass filter from 0 to 700 cps, then through a
high pass filter from 3500 to above 10,000 aps, and finally, through
these two filters in combination. It was found that the noise
power could be raised from 34 to 27 db above the maximum speech
power before intelligence was considerably reduced. The relative
levels were measured on a root mean square level recorder, and also
on an oscilloscope, Next, noise was introduced in the speech band.
For this case the noise power level had to be reduced I. db below
Sanitized Copy Approved for Release 2011/05/03: CIA-RDP78-0330OA001900110027-9
 Sanitized Copy Approved for Release 2011/05/03: CIA-RDP78-03300A001900110027-9
CONFIDENTIAL
2.
the speech level for good intelligibility. For wide band signals,
good intelligence is obtained when the noise and speech levels are
about equal.
Third, a limited number of selected words were chosen which
have relatively flat formants; i.e., formants whose frequencies are
constant. These words were analyzed on a sound spectrograph, and
the frequency regions of maximum energy were obtained. These
speech samples were then passed through a set of three bandpass
filters which passed the frequencies of maximum speech energy.
Subjective listening tests were carried on to determine the value
of this type of noise discrimination. No conclusive results were
obtained from this experiment. It appears that subjective listening
tests will not provide a satisfactory criterion when the tests are
made on narrow band information. Some type of objective measure-
ment must be devised which will give a more definite result.
As yet, no definite conclusion on bandwidth or band spacing
have been reached.
September 6, 1956
CONFIDENTIAL
Sanitized Copy Approved for Release 2011/05/03: CIA-RDP78-03300A001900110027-9