Charles River Analytics was awarded a Phase II Small Business Innovation Research (SBIR) contract from the Naval Air Systems Command Naval Air Warfare Center Training Systems Division (NAWCTSD). Supported by this funding, the Charles River team and partner OWT Global are developing techniques to tailor speech processing to specialized training applications.
Speech recognition and generation are advanced in consumer electronics, such as virtual assistants. However, speech processing is still challenging in environments such as defense training programs with its specialized vocabulary and grammar. Commercial speech understanding systems have trouble with military language, which contains idiosyncratic vocabulary and grammar. Generating speech for training is also a challenge.
For example, air traffic control trainees use recorded speech to practice identifying important communication in a noisy background, but trainees quickly learn to recognize and ignore the recordings, which gives them an unfair advantage and makes the training less effective.
To solve these issues, Charles River Analytics is using novel natural language processing (NLP) techniques, including developing custom speech models for understanding, and automatically generating novel speech on the fly. Communications with Operational Context and Knowledge for Target Audio Identification Learning (COCKTAIL) involves applying these techniques to help train air traffic controllers to communicate effectively in a noisy environment.
“When we improve the techniques for using speech processing technology to understand and generate communication for training applications, it will have wide applicability across the military because of how central training is to their work,” says Dr. Terry Patten, Principal Scientist at Charles River Analytics and Principal Investigator of the COCKTAIL effort.
COCKTAIL will immediately support target audio identification training in the Navy, Air Force, and the Federal Aviation Administration. Furthermore, it could enhance Charles River’s KWYN™ suite of training solutions and commercial applications that require speech processing in training systems.
This material is based upon work supported by the Naval Air Systems Command (NAVAIR) under Contract No. N68335-23-C-0300. Any opinions, findings and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of NAVAIR.