Individualized, Noninvasive Speech Indicators for Tracking Elevations in Stress (INSITES)

The complexity and round-the-clock nature of NASA operations in low Earth orbit (LEO) and future cis-lunar deep space missions, along with isolation in the extremely hostile environment of space, can induce levels of acute and chronic stress that could compromise astronaut performance, leading to errors that could affect science payloads, crew safety and mission success. For the exploration of space, therefore, a method is needed to assess operator state, quickly and reliably detect stress, and provide objective feedback to the individual, crew, and ground support, in order to mitigate adverse events and mishaps. Due to the unique challenges of NASA missions, Quantum Applied Science and Research (QUASAR) proposes to develop a system to identify Individualized, Noninvasive Speech Indicators for Tracking Elevations in Stress (INSITES). The overall goal of this INSITES project is to develop an unobtrusive, objective, and reliable detector of stress that measures changes in speech and vocalizations using equipment that would be present or used (microphones, communications systems, computers) used during operations, thus not requiring additional sensors or dedicated processing hardware. QUASAR and the Florida Institute for Human and Machine Cognition (IHMC) will build a database of audio stress recordings acquired under laboratory conditions in order to construct normative models of stress, using vocal stress-related features identified in Phase I. A methodology for recalibrating normative models to individuals using minimal additional training data in order to optimize model performance will be developed. QUASAR will also prepare and validate an INSITES prototype that will provide a real-time visual output describing an individualrsquo;s stress level. The prototype will be based upon an app that can be readily installed on a mobile device or implemented in NASA spacecraft and habitats to detect changes in stress acutely and over time.