You are here
Speech Interface Architecture For Human To Agent Interactions
Title: Principal Investigator
Phone: (215) 542-1400
Email: jstokes@chiinc.com
Title: Mgr, Contracts
Phone: (215) 542-1400
Email: prollhauser@chiinc.com
Contact: Max Louwerse
Address:
Phone: (901) 678-2143
Type: Nonprofit College or University
"The use of agents as synthetic entities is a key technology in the development of next generation simulation based training and performance support systems. In these systems agents must collaborate with humans, communicating through speech. Unlike manyexisting speech technology applications which require only simple sequential dialogs, military domains frequently involve time critical communications. This fact imposes requirements for real time operation and the support of more complex speechinteractions, such as interruptions. Unfortunately many of the technologies available for system development impose latencies on the processing of speech interaction. This is particularly problematic in the case of speech understanding, which can becomecomputationally intensive when robust performance is required, as for fluid human-agent speech interaction. We propose to analyze the available technology solutions, including network solutions commonly used in virtual environments, identify latencies,their impact, and possible solutions. Available speech understanding approaches will be evaluated and selected for real time viability. General networking solutions based on compensation, circumvention, and prioritization will be developed, resulting ina network speech protocol for human-agent speech interaction. Captured in a human-agent reference architecture, the Phase I results will drive the Phase II development of Voice Interaction and Recognition for Virtual Environments - VIRVE. The proposedresearch will lead to the development of technology that can be employed in a broad variety of military and commercial simulation based training and performance support systems. A fully developed reference architecture, network speech protocol, and realtime speech understanding components will support the construction of agent-based synthetic team members for training and collaborative task performance. These technologies will be of considerable value in domains such as emergency response and disasterrelief, as well as a wider range of military sett
* Information listed above is at the time of submission. *