DYNAMIC PAGE -- HIGHEST POSSIBLE CLASSIFICATION IS
TOP SECRET // SI / TK // REL TO USA AUS CAN GBR NZL

(U//FOUO) NSA Georgia Opens New Audio-Forensics Lab
FROM: Marti Hix Mercer
Chief, Corporate Communications, NSAG
Run Date: 02/22/2007

(S//SI//REL) New lab aims to help analysts fully capitalize on intercepted voice communications
from our targets.
(S//SI//REL) The start of the new year also saw the start of a powerful new capability at NSAG - a new audio-forensics lab funded through the Senior Language Technology Authority (SLTA).
The lab -- the first of its kind anywhere within the Extended Cryptologic Enterprise -- initially will
serve NSAG, with a vision toward expanding to accomodate external requests. It aims to help
solve critical language processing problems confronting Language Analysts. As envisioned, the
lab will provide dedicated, on-site, state-of-the-art solutions to address the problems that impair
the processing of critical foreign language voice intercept.
(S//SI//REL) The intent is to provide a capability to language analysts as quickly as possible and
support them in the evaluation of that capability. According to 
 the lab's audio
engineer, the lab is in the "early walking" phase of a crawl-walk-run approach. As the audio
forensics lab at NSAG succeeds, the SLTA hopes to build similar labs at the other major regional
cryptologic centers.

(S//SI//REL)

 in the audio-forensics lab

(TS//SI//REL
 directly supports NSAG language analysts by enhancing hearability and
intelligibility of audio intercepts. He does this by performing audio restoration or audio forensic
analysis to help language analysts determine the content and context of high-priority GWOT*related audio intercepts. 
lso works on fundamental problems associated with speech,
speaker identification, age and gender determination, psychological stress in voice, speech
intelligibility/comprehension, speech signal enhancement, and masking the effects of noise on
speech. Specific goals of the lab are to:
Filter and enhance unintelligible/poor-quality voice intercept;
Assist NSAG language analysts or tactical forces with voice/speaker identification to aid in
recognition of high value individuals or detainees;
Identify, isolate, and extract background speakers and audio sounds;
Develop signal databases containing speaker, context, background, and other information
with the ability to index, sort, and retrieve in real time;
Collaborate in the analysis and processing of complex voice intercept.
(S//SI//REL) The audio lab provides an unparalleled opportunity to support prototype testing
and spiral development. In addition, the lab acts as a test-bed for language processing
capabilities developed elsewhere. The lab will partner with the SLTA, the Research Directorate,
the Human Language Technology Program Management Office, and Speech Services -- the
providers of operational speech technologies ( BABBLEQUEST ) to customers across the
Enterprise -- to test capabilities developed from these organizations before placing them in an
operational environment.
(U//FOUO) The audio lab has an open architecture so that it can readily accept new capabilities,
tools, and technologies as they are developed. In addition, the lab will be net-centric so that

 language analysts will be able to collaborate with the lab virtually. Further, if and when similar
audio labs are set up elsewhere throughout the SIGINT Enterprise, they can be networked
together to leverage and utilize existing capabilities.
(U//FOUO) Finally, in addition to being a fully functioning audio forensics lab, the room itself is
designed to be a showcase for NSAG. Consequently, it reflects an impressive, high-tech, visually
appealing environment that is pleasant to work in.
is very much looking forward to
working with language experts throughout the Enterprise in order to increase the lab's capacity
to improve the Agency's capabilities. As a plaque in the lab states, "Supporting the mission
never sounded so good."

(S//SI//REL) At work in the lab
(U) Notes: 
* (U) GWOT = Global War on Terrorism

"(U//FOUO) SIDtoday articles may not be republished or reposted outside NSANet
without the consent of S0121 (DL sid comms)."

DYNAMIC PAGE -- HIGHEST POSSIBLE CLASSIFICATION IS
TOP SECRET // SI / TK // REL TO USA AUS CAN GBR NZL
DERIVED FROM: NSA/CSSM 1-52, DATED 08 JAN 2007 DECLASSIFY ON: 20320108