Neuromorphic and Digital Hybrid Systems

--Neuromorphic systems have been primarily implemented using analog sub-threshold CMOS circuits primarily because the physics of subthreshold operation and biological signal processing are identical. The effects of attempting to model biology accurately, coupled with the slow dynamics of sub-threshold MOS circuits, have prevented these systems from capitalizing on the main benefit of silicon which is its speed. Therefore the full potential of the silicon medium will be realized when neuromorphic systems are implemented with mixed signal circuits, where the essence of the biological computation is implemented with circuits which are best suited for the task. Presented are three hybrid systems, which realize the philosophy that the optimal neuromorphic VLSI circuit is one which imitates biology only when it is advantageous, uses sub-threshold circuits when speed is not required, and otherwise uses mixed signal strong inversion circuits.

[1]  Eric A. Vittoz,et al.  Analog VLSI implementation of neural networks , 1990, IEEE International Symposium on Circuits and Systems.

[2]  Dean A. Pomerleau,et al.  Vision guided lane transition , 1995, Proceedings of the Intelligent Vehicles '95. Symposium.

[3]  M. Alexander,et al.  Principles of Neural Science , 1981 .

[4]  Christof Koch,et al.  Vision Chips: Implementing Vision Algorithms with Analog VLSI Circuits , 1994 .

[5]  Jan Van der Spiegel,et al.  A focal plane visual motion measurement sensor , 1997 .

[6]  D. Robinson.  The mechanics of human smooth pursuit eye movement. , 1965, The Journal of physiology.

[7]  Y. J. Tejwani,et al.  Robot vision , 1989, IEEE International Symposium on Circuits and Systems,.

[8]  Cesar Bandera,et al.  Foveal machine vision systems , 1989, Conference Proceedings., IEEE International Conference on Systems, Man and Cybernetics.

[9]  Christof Koch,et al.  Analog VLSI Circuits for Attention-Based, Visual Tracking , 1996, NIPS.

[10]  W. Reichardt,et al.  Autocorrelation, a principle for the evaluation of sensory information by the central nervous system , 1961 .

[11]  Christof Koch.  Implementing early vision algorithms in analog hardware: an overview , 1991, Defense, Security, and Sensing.

[12]  Mona E. Zaghloul,et al.  Silicon Implementation of Pulse Coded Neural Networks , 1994 .

[13]  Michael P. Eckert,et al.  Effect of tracking strategies on the velocity structure of two-dimensional image sequences , 1993 .

[14]  J. Van der Spiegel,et al.  A Foveated Visual Tracking Chip , 1997, 1997 IEEE International Solids-State Circuits Conference. Digest of Technical Papers.

[15]  Kwabena Boahen.  Retinomorphic Vision Systems 11: Communication Channel Design , 1996, 1996 IEEE International Symposium on Circuits and Systems. Circuits and Systems Connecting the World. ISCAS 96.

[16]  J. G. Elias.  Silicon Dendritic Trees , 1994 .

[17]  Ralph Etienne-Cummings,et al.  Accurate and simple CMOS 'one-over' circuit , 1993 .

[18]  D. Sparks,et al.  Population coding of the direction, amplitude, and velocity of saccadic eye movements by neurons in the superior colliculus. , 1990, Cold Spring Harbor symposia on quantitative biology.

[19]  R. Bajcsy.  Active perception , 1988 .

[20]  白井 良明.  MITのArtificial Intelligence Laboratory , 1973 .