Crew resource management training systems are often constrained by the high cost and lack of flexibility in coordinating a large group of human role players for part-task training. Motivated by the recent maturation of the speech synthesis and recognition technologies, speech-enabled crew role-player agents are being introduced to address these limitations. However, difficulties remain in customizing the synthetic speech agents in a timely manner. This proposal advances the development of speech agent implementation tools that will allow the end-users such as instructors and scenario designers to readily customize the set of supported utterances without external support and to provide pre-training modules for the trainees for enhancing speech recognition accuracy. Proposed work will leverage technologies from previous speech recognition technology development projects with the Navy and NASA. Feasibility demonstration of the proposed tool suite in the Anti-Submarine Warfare domain was carried out during the Phase I research. Extended from Phase I, Phase II will develop, demonstrate, and validate a prototype system. The Phase II deliverable will include a working prototype of the software that demonstrates the capability of enhancing utilization of synthetic speech agents in the Navy aviation training environment. Commercialization activities will be undertaken during the Phase III work.
Benefit: The proposed research effort will create a software suite that enables rapid domain adaptation of an automatic speech recognition systems used in crew resource management systems. It will also be capable of providing end-user training for optimizing the interactions with the speech recognition system. Capabilities developed under the proposed research can be used for military and civilian mission training. For example, the developed technology will directly benefit Naval aviation training systems. These systems are also useful for training teams of individuals in critical applications such as civil disaster management.
Keywords: Anti-Submarine Warfare, speech understanding, Language Modeling, Automatic Speech Recognition, Natural Language Processing, Crew Resource Management Training, Domain Adaptation, speech system training