Isolating and Locating Speakers in Clutter

The challenge is already upon us that our military forces have to perform operations in urban environments. Our adversaries are studying our strategies, doctrine and technologies. While they may not be able to meet these capabilities in a time of conflict, they are doing what ever necessary to avoid operational environments for which our forces are optimized. With this in mind, they are seeking complex urban terrain for cover and concealment to try and offset our superior capabilities. Cities have become primary sanctuaries for many contemporary insurgents and therefore require a solution to ascertain and disseminate information about the persons "hiding" in those environments with the ability to remotely monitor their conversations. Improvements to algorithms that have already shown robust speech recognition intelligibility and audio tracking can be adapted to perform specific speaker extraction, identification and recognition in extremely high noise and chaotic aural cluttered environments. With this information, our forces will have a unique HUMINT capability to separate and identify persons of interest based on unique spoken utterance characteristics. Captured audio data can be used to track their movements, identify plans or potential threats based upon the content of the conversation, and perform speaker identification and verification type tasks.