Our research covers analog and digital VLSI microsystems for adaptive neural computation and sensory information processing, from neuromorphic systems engineering and kernel-based learning machines to micropower implantable neural interfaces, acoustic microarrays, adaptive optics and biometric identification. Our group contributed a variety of highly efficient analog VLSI processors for vision, audition and pattern classification which outperform general-purpose digital processors with a factor 100-10,000 reduction in power dissipation and similar savings in silicon area.
The focus of our current research is on cross-cutting advances at the interface between in vivo and in silico neural information processing. The goals are threefold: to empower silicon integrated circuits with adaptive intelligence inspired by sensory information processing in nervous systems; to facilitate advances in computational neuroscience by large-scale emulation of neural models in parallel analog silicon circuits; and to interface silicon with neural cells for restoring lost function in sensory and motor impaired patients.
Our work continues to pursue advances in the cognitive performance and efficiency of neuromorphic cognitive adaptive systems and wireless brain interfaces.
Texas Instruments, G. Cauwenberghs, 6/2012-5/2015.
Adaptation, learning and memory in analog VLSI have been the focus of extensive research in our laboratory. The edited volume Learning on Silicon (Kluwer Academic, 1999) serves as a reference to this emerging field. Our work formulated and implemented mechanisms of adaptation and learning embedded in parallel distributed architecture which, like synaptic plasticity in nerve assemblies, sustain computational intelligence in variable and complex environments and compensate for imprecision in analog computation. This research earned the recognition of a Presidential Early Career Award for Scientists and Engineers (PECASE) and an ONR Young Investigator Award. Our work extended to kernel learning machines which incorporate principles of statistical learning theory. Notably we developed the Kerneltron, a support vector "machine" as a massively parallel VLSI array processor on a single chip with energy efficiency on par with synaptic transmission in the mammalian brain.
Qualcomm Inc., G. Cauwenberghs, 5/2010-4/2012.
National Semiconductor, G. Cauwenberghs, 11/2009-11/2010.
Design and implementation of a low-weight wearable, wireless EEG recording system for non-intrusive acquisition and on-line interpretation of brain dynamics, supporting the study of learning in dynamic environments with freely interacting subjects.
National Institute of Aging, NIH/HIA 1R01AG029681, G. Cauwenberghs, D. Kleinfeld, T. Sejnowski and N. Thakor, 9/06-6/10.
Development and application of high-resolution functional imaging techniques to study the interaction between bloodflow and neural activity in cortex at micrometer and millisecond resolution.
DARPA, T.P. Jung, S. Makeig, G. Cauwenberghs, I. Galton, and T. Sejnowski, 6/06-6/07.
Design and implementation of wireless instrumentation for high-resolution EEG functional brain imaging, using dry electrode sensing technology and on-line independent component analysis.
DARPA subcontract, G. Cauwenberghs and P. Yu, with M. Vorontsov (Army Research Laboratory) and T. Dorschner (Raytheon), 6/06-11/08.
Investigation into high-resolution microscale adaptive optics and stochastic adaptive control for aberration correction, implemented in parallel analog VLSI.
National Science Foundation, R. Etienne-Cummings and G. Cauwenberghs, 10/04-9/07.
Design, analysis and implementation of vision algorithms onto focal-plane image processors, for surveillance of the visual scene.
Catalyst Foundation Award, G. Cauwenberghs, L. Degertekin, and G. Zweig, 9/03-8/07.
Design and implementation of an integrated MEMS/VLSI optical microphone array and signal processor to extract multiple sound sources from the acoustic environment.
Projects towards the development of a custom-trainable, versatile, self-contained and mobile system for visually impaired users. The system will aid the user in interacting freely with other people and the environment, by rapidly detecting and localizing key visual environmental cues, and rapidly recognizing and identifying familiar people and objects. At the core of the system is the Kerneltron, a massively parallel Support Vector "Machine" (SVM) in silicon.
Defense Advanced Research Projects Agency and Office of Naval Research, G. Cauwenberghs, A.G. Andreou and R. Etienne-Cummings, Subcontract to University of Maryland (S. Shamma), 9/00-8/03.
In this collaborative project with the University of Maryland, University of Syndey, and Signal Systems Corporation, our goals are to design and demonstrate novel MEMS acoustic sensor arrays, and to integrate these MEMS sensors with adaptive VLSI systems for real-time applications of signal tracking, source identification and active noise control where micropower and miniature operation are a strict requirement.
Office of Naval Research Young Investigator Award and Presidential Early Career Award for Scientist and Engineers (PECASE), ONR #N00014-99-1-0612, G. Cauwenberghs, 3/99-3/04.
An investigation in the problem of separating mixtures of signals when nothing or little is known about the sources of the signals or the way they are mixed. Independent Component Analysis provides a mathematical basis for some of this work. The objectives are to retrieve the sources from the mixture(s) and to identify their spatial origin. The goal is to develop novel algorithms, and prototype integrated processors that implement the algorithms in real time for portable applications in speech, acoustics and sonar signal processing.
Office of Naval Research, ONR #N00014-99-1-0654, G. Cauwenberghs, R.T. Edwards and F. Pineda, 4/99-9/00.
A collaborative project with Orincon to develop an integrated signal processor for real-time classification of underwater buried targets from active sonar inspired by dolphin and bat echolocation. The goal is a miniaturized low-power system suitable for use in an autonomous underwater vehicle.
An interdisciplinary program of research and education in custom integrated computational systems, concentrating on analog VLSI systems for multimedia sensory integration, covering applications of speech processing, visual motion estimation, and communications.
Grant supporting Ph.D. research on interdisciplinary research involving circuit design. Low-power mixed analog-digital VLSI implementation of a hybrid ANN/HMM (artificial neural network/hidden markov model) system for continuous speech recognition.
Multidisciplinary University Research Initiative (MURI) (ONR/DARPA N00014-95-1-0409), A. Andreou and G. Cauwenberghs, with Boston University (S. Grossberg et al.), 6/96-5/00.
A multidisciplinary research initiative (MURI) center for automated vision and sensing systems, with applications to synthetic aperture radar (SAR) image processing and pattern recognition, and automated active vision systems with sensory-motor integration. The Johns Hopkins component of the project contributes dedicated analog VLSI sensory information processing systems as key components of active vision systems, and as dedicated co-processors interfacing with digital computers.
Maryland Industrial Partership Program (MIPS) and Northrop Grumman, G. Cauwenberghs, 2/97-2/99.
Design and VLSI prototyping of high-speed oversampling and pipelined analog-to-digital converters in CMOS, BiCMOS and SiGe technology. With the Systems Development and Technology division of Northrop Grumman.
Army Research Laboratory (ARO/Battelle), G. Cauwenberghs, 2/97-1/98.<
A seed project to interface analog VLSI electronics with optical diffraction systems for on-line correction of abberation of optical propagation in atmospheric media. In collaboration with the Intelligent Optics Laboratory at ARL.
Army CECOM/DCS Corporation, A. Andreou and G. Cauwenberghs, 9/94-11/98.
Development of hybridized focal-plane array IR sensors, interfacing HgCdTe arrays and other IR-responsive materials with silicon circuits through flip bonding. VLSI algorithms for on-line offset correction. Joint work with Army CECOM Fort Belvoir.
National Science Foundation, Collaborative Research Initiative (CRI), NSF IBN-9634357, E. Niebur, M. Steinmetz and G. Cauwenberghs, 2/97-1/98.
An interdisciplinary project combining theoretical and experimental neuroscience with analog VLSI emulation of mechanisms of attention in biological neural systems. In collaboration with and coordinated by the Mind-Brain Institute.<