Compressive sampling by artificial neural networks for video

Harold Szu; Charles Hsu; Jeffrey Jenkins; Kitt Reinhardt

doi:10.1117/12.887596

8 June 2011 Compressive sampling by artificial neural networks for video

Harold Szu, Charles Hsu, Jeffrey Jenkins, Kitt Reinhardt

Proceedings Volume 8058, Independent Component Analyses, Wavelets, Neural Networks, Biosystems, and Nanoengineering IX; 805803 (2011) https://doi.org/10.1117/12.887596
Event: SPIE Defense, Security, and Sensing, 2011, Orlando, Florida, United States

Abstract

We describe a smart surveillance strategy for handling novelty changes. Current sensors seem to keep all, redundant or not. The Human Visual System's Hubel-Wiesel (wavelet) edge detection mechanism pays attention to changes in movement, which naturally produce organized sparseness because a stagnant edge is not reported to the brain's visual cortex by retinal neurons. Sparseness is defined as an ordered set of ones (movement or not) relative to zeros that could be pseudo-orthogonal among themselves; then suited for fault tolerant storage and retrieval by means of Associative Memory (AM). The firing is sparse at the change locations. Unlike purely random sparse masks adopted in medical Compressive Sensing, these organized ones have an additional benefit of using the image changes to make retrievable graphical indexes. We coined this organized sparseness as Compressive Sampling; sensing but skipping over redundancy without altering the original image. Thus, we turn illustrate with video the survival tactics which animals that roam the Earth use daily. They acquire nothing but the space-time changes that are important to satisfy specific prey-predator relationships. We have noticed a similarity between the mathematical Compressive Sensing and this biological mechanism used for survival. We have designed a hardware implementation of the Human Visual System's Compressive Sampling scheme. To speed up further, our mixedsignal circuit design of frame differencing is built in on-chip processing hardware. A CMOS trans-conductance amplifier is designed here to generate a linear current output using a pair of differential input voltages from 2 photon detectors for change detection---one for the previous value and the other the subsequent value, ("write" synaptic weight by Hebbian outer products; "read" by inner product & pt. NL threshold) to localize and track the threat targets.

Citation Download Citation

Harold Szu, Charles Hsu, Jeffrey Jenkins, and Kitt Reinhardt "Compressive sampling by artificial neural networks for video", Proc. SPIE 8058, Independent Component Analyses, Wavelets, Neural Networks, Biosystems, and Nanoengineering IX, 805803 (8 June 2011); https://doi.org/10.1117/12.887596

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available