Second-Order Neural Core for Bioinspired Focal-Plane Dynamic Image Processing in CMOS

Abstract

Based on studies of the mammalian retina, a bioinspired model for mixed-signal array processing has been implemented on silicon. This model mimics the way in which images are processed at the front-end of natural visual pathways, by means of programmable complex spatio-temporal dynamic. When embedded into a focal-plane processing chip, such a model allows for online parallel filtering of the captured image; the outcome of such processing can be used to develop control feedback actions to adapt the response of photoreceptors to local image features. Beyond simple resistive grid filtering, it is possible to program other spatio-temporal processing operators into the model core, such as nonlinear and anisotropic diffusion, among others. This paper presents analog and mixed-signal very large-scale integration building blocks to implement this model, and illustrates their operation through experimental results taken from a prototype chip fabricated in a 0.5-m CMOS technology.

An eight layer cellular neural network for spatio‐temporal image filtering

“…This work has focused on the latter, where we have increased architectural complexity by increasing the number of layers and by introducing non-linear coupling between layers via the output non-linearity. This work follows a recent trend towards the development of multi-layer CNN architectures [26,31,32], which has been motivated in part by the observation that biological structures for visual processing, such as the retina, also exhibit a multi-layered architecture [33][34][35]. Table III summarizes important parameters of several recent VLSI implementations of multilayer CNN architectures, as well as a reference single layer design, the ACE16K [30].…”

Learning of spatiotemporal behaviour in cellular neural networks

“…Moreover, these equations describe closely the model of the CACE1k chip [18] which is nowadays a major advance in VLSI implementation for generation of complex behaviour. Locality and invariance are the main advantages of CNNs.…”

“…This equation can also be written in its condensed form: These equations describe the behaviour of a second-order cell, which when isolated from the neighbouring cells may behave as an oscillator [17] and when coupled to other cells is able to generate interesting spatiotemporal patterns. Moreover, these equations describe closely the model of the CACE1k chip [18] which is nowadays a major advance in VLSI implementation for generation of complex behaviour. Locality and invariance are the main advantages of CNNs.…”

Sensor Integration in Autonomous Systems

“…In the process of building these models, the importance of interactions between multiple layers or arrays became apparent. A custom VLSI chip was designed and implemented by identifying a basic multi-layer interaction required to implement these models [16]. Once this chip was available, it could be programmed to implement the desired models [17].…”