Enabling Rapid Human Knowledge Encoding via Feature-Based Neurotechnology

Abstract

The aim of the proposed work is to successfully read-out feature-based processing layers (FPLs) of the human visual system, using cheap non-invasive consumer-grade technology. If this fundamental research proposal is successful, it will serve as a basis to develop a future Rapid Neurotechnology-Assisted Learning proof-of-concept system as well as other warfighter and defense applications derived from successful demonstration of this fundamental neurotechnological capability. Decoding of perception and cognitive processes, occurring in the FPLs of the brain, are the missing technological capability for true human-machine coupled learning applications. This is an ambitious advance beyond previous work that have generally focused on pulling neurocorrelates of (1) general engagement/task readiness, or (2) target selection or target detection, or (3) Brain-Computer-Interface (BCI) intentional motor activity capture. Project Technical Aims 1. Achieve significant ROC for FPLs of early perception (e.g. familiarity, visual complexity) 2. Achieve significant AUC ROC for FPLs of stimuli context (e.g. position, temporal) 3. Achieve significant AUC ROC for FPLs of conceptual grouping (e.g. absolute, relative) 4. Achieve at least a 50% accuracy on group identity (out of at least a 5-class set) 5. Demonstrate at least one intermediate for each FPL with at least a 0.7 AUC ROC The goal of the proposal is to demonstrate the accurate capture of multiple FPLs under passive conditions where the task goal is not explicitly set Ð not under active task conditions (explicit participant instruction). In most prior projects, the task goal is explicit and identical to the neural signal target. In order for the FPL signal to be valid, we will have to demonstrate the signal does not have an active task correlation. Additionally, in order for the FPL signal to be useable in the future by a learning technology, the signal will have to be capturable in a rapid visual presentation setting. Finally, we will have to achieve this using technologies that are not cost-prohibitive or user-effort-prohibitive. Towards that end, we will develop appropriate analytical and hardware modifications to capture these FPLs.

Document Details

Document Type
DoD Grant Award
Publication Date
Jul 09, 2020
Source ID
W911NF2010028

Entities

People

  • Omar Claflin

Organizations

  • Army Contracting Command
  • Defense Advanced Research Projects Agency

Tags

Readers

  • Agent-Based Social Robotics and Mobile-Assisted Learning in Virtual Environments.
  • Forest Ecology
  • Sensor Fusion and Tracking Systems.