The intelligence community is interested in extracting tactical situation knowledge of current planned enemy activities from the automated monitoring of communications signal traffic by isolating and tracking a speakers voice and linguistic context. The limiting factor in the successful application of these emerging language based interpreters is the poor quality of the speech captured from these intercepted transmissions. NLU, language identification, language patterns, speaker identification, keyword spotting, gisting, language parsing and context interpretation are emerging technologies to be matured to achieve this automated intelligence processing. Unique to this approach is the advanced research of innovative wide-band signal recovery techniques for accurate extraction of speech from the severely degraded transmission intercepts, using natural language information extraction methods that target terrorist activities. Unique to this research opportunity is the ability for us to merge our recent technology breakthroughs for identifying and tracking individual speakers and automated filtering to search for specific semantic and language style constructs. We have demonstrated highly accurate speech extraction algorithms in the presence of very high noise environments with a SNR of minus 10dB. These previously developed signal recovery and speech extraction algorithms are designed to work as a front-end processor for speaker identity, recognition and interpretation algorithms. We are currently developing and integrating a superior NLU interpretation and understanding system for information extraction and activity gisting using syntactic, semantic and patterned dialogue context understanding algorithms. The goal of this research is to investigate and develop novel techniques for the deconvolution of specific communication channels from the total spectrum based upon learned terrorist characteristic feature content. Maximum likelihood statistics, neural network and signal separation technologies have progressed sufficiently to support the development of algorithms for the extraction of these signals of interest. This research will result in software algorithms, suitable for execution in future Army tactical and strategic based communication assets, which perform the intelligence gathering functions. Developed software modules will be portable and reusable as well as compatible with the Joint Technical Architecture-Army tactical and commercial wireless architectures. This technology has commercial application to law enforcement investigations in which the exploitation of cell phone and other wireless transmissions become necessary, as in cases where threatening phone calls are being made or in surveillance missions. Automated natural language interpretation and understanding processing represents the preferred next generation interface between humans and their supporting electronic devices. Improved speech and language intelligibility processing will provide wider user acceptability of the automated natural language technology in a wide range of consumer products and services, including personal data assistants, computers, controls and an intuitive interface for appliances, radios, cell-phones, vendor kiosks, web-site browsers, information fusion and decision aids
