MOBBED (Mobile Brain-Body-Environment Decisionmaking) Part I: Database Design
Date
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
The US Army CAN-CTA project focuses on understanding the brain and body in action using a data-driven approach with a goal of developing assistive and monitoring technologies for soldiers in real world environments. Project scientists collect EEG, eye-tracking, motion capture, muscle activity, heart rate, force-plate responses, as well as sensory and galvanic skin responses from subjects performing a variety of tasks in realistic environments. Video and audio recordings, both of subjects and of the environment are also available. Some datasets, acquired in controlled simulation environments, have detailed simulation-specific information that describes and controls the context. Some subjects will undergo additional imaging modalities (such as diffusion tensor imaging or simultaneous fMRI and EEG recording) in fixed environments to obtain detailed structural brain information.
Brain-body imaging in realistic environments has only recently become feasible with the advent of portable dry-electrode technology. Various partners within the CAN-CTA are beginning to acquire datasets that combine various modalities. However, the equipment, data format, processing, and storage techniques vary considerably across groups. The CAN-CTA data corpus will be, for the most part, annotated streaming data characterized by diverse formats, high data rates, and complex interrelationships. Analysis and discovery in such a data corpus requires a data handling structure that is robust, extensible, and capable of extracting complex combinations of features for analysis and visualization.
Two fundamental assumptions have been widely articulated among CAN-CTA members. The first assumption is that effective data-driven approaches require a data corpus. Without training and test data that covers the range of possibilities, machine learning and other computational approaches fail. The second assumption is that the corpus of data acquired under the initial CAN-CTA can provide a lasting legacy with far-reaching potential beyond the individual experiments and particular research questions that initiated these experiments.