Project Title:
Hybrid Processor for Physiological Artifact Detection
12.03-3474
911341
Hybrid Processor for Physiological Artifact Detection
Charles River Analytics, Inc.
55 Wheeler Street
Cambridge
MA
02138
Greg L.
Zacharias
617-491-3474
ARC
NAS2-13532
242
12.03-3474
911341
Abstract:
Hybrid Processor for Physiological Artifact Detection
Phase I will evaluate the feasibility of developing a hybrid physiological artifact
detection system from two complementary artificial intelligence (AI) technologies:
artificial neural networks (ANNs) and knowledge-based expert systems (ESs). By hybridizing
these two technologies, computer-based pattern recognition (via ANNs) will be combined
with a human expert's knowledge of physiological signal recording characteristics
(via ESs). This hybrid detector will be developed within the existing NueX software
development environment which supports synergistic interaction between ANNs and kbESs.
The envisioned prototype artifact detection processor will support dynamic updating
of the physiological signal knowledge base, via the artifact recognition capabilities
of the ANN, and result in continuous learning by the ANN, via the recording paradigm
knowledge stored in the ES. Feasibility will be evaluated by defining the scope of
the problem and identifying candidate solutions; designing and implementing a prototype
hybrid detector; demonstrating and evaluating detector performance; and specifying
hardware and software system design requirements for Phase II implementation of a
real-time system.
Commercial potential exists for the end product itself, a generic physiological artifact
detector, and for the hybrid software environment used to develop it. The hybrid
processor holds promise for inclusion in a wide range of existing physiological instrumentation
systems. The development software can serve as the basis of such other signal detection
and/or isolation applications as exist in the area of fault detection and/or isolation
and safety monitoring.
physiological signal processing, artificial neural networks, expert systems