Products that implement the ROSAA (Reconfigurable, Open-Systems, Avionics Architecture)

The intelligence community is interested in extracting tactical situation knowledgeof current and planned enemy activities from the automated monitoring of battlefieldradio communications. The limiting factor in successful application of theseemerging language based interpreters is the poor quality of the speech captured fromthese tactical radio intercepted transmissions. Recovery of the original qualityspeech responds to advanced signal recovery techniques, applying complex adaptivespline neural network recovery in combination with blind separation and speechdetection methods. Similarly, the intelligibility of longhaul reach-backcommunications is dependent upon the same signal recovery technology as applied tomultiple link scenarios from the battlefield to CONUS. Natural LanguageUnderstanding (NLU), language identification, language translation, speakeridentification, keyword spotting, gisting, language parsing and contextinterpretation are emerging technologies to be developed to achieve this automatedintelligence processing. Innovative wideband signal recovery techniques, are usedfor accurate extraction of captured speech from the severely degraded transmissionintercepts, incorporating methods most beneficial for natural language algorithmperformance and to discount radio specific signatures. Unique to this researchopportunity is the ability to merge our recent technology breakthroughs foridentifying and tracking speakers by stripping away radio channel masking andmultipath degraded tactical radio transmitted speech. TRI has demonstrated highlyaccurate speech extraction algorithms are possible in noise environments with aSignal-to-Noise Ratio (SNR) of minus 10dB. These TRI emerging signal recoveryspeech extraction algorithms are designed to work as a front-end processor for NLUrecognition and interpretation algorithms. TRI is also developing and integrating asuperior speech interpretation and understanding system for information extractionfrom tactical transmissions using syntactic, semantic and tactical dialog contextunderstanding algorithms. The real-time speech intelligibility enhancementtechnology is to be embedded with the next generation radio audio processing schemato effect a processing efficient and low-power consumption solution by mergingduplicative speech processing and memory requirements.