this article we present a hardware model of a selective attention mechanism implemented on a very largescale integration (VLSI) chip, using analog neuromorphic circuits. The chip exploits a spike-based representation to receive, process, and transmit signals. It can be used as a transceiver module for building multichip neuromorphic vision systems. We describe the circuits that carry out the main processing stages of the selective attention mechanism and provide experimental data for each circuit. We demonstrate the expected behavior of the model at the system level by stimulating the chip with both arti#cially generated control signals and signals obtained from a saliency map, computed from an image containing several salient features

This paper presents a one-dimensional visual sensor, implemented on a single VLSI chip using analog neuromorphic circuits, for selectively detecting and tracking the position of the feature with the highest spatial contrast present in the visual scene. The chip's photoreceptors adapt to stationary backgrounds and can be tuned to respond maximally to specific target velocities. The sensor drastically reduces the amount of data to be transmitted to further processing stages by encoding, in real time, the position of the target in the form of a single continuous-time analog variable. We describe the circuits implementing the sensor and show applications to three examples of tracking tasks: a stand-alone visual tracking system, an active fully analog tracking system, and a mobile platform line-following system.

Winner-take-all (WTA) circuits are commonly used in a wide variety of applications. One of the most used current-mode WTA designs is the one originally proposed by Lazzaro et al. [1]. Several extensions to this design have been suggested in the past. In this paper we present a variant of this current-mode WTA circuit, containing all of the enhancements previously proposed, together with new additional modifications that endow it with interesting hysteretic and lateral inhibition and excitation properties. We compare the performance of this WTA circuit to the original WTA design, providing experimental data obtained from a VLSI chip containing both types of circuits, designed using closely matched layouts. We derive analytically the response properties of the circuit's lateral diffusion network, pointing out the differences to previously proposed diffusion networks, and present experimental data confirming the theoretical predictions. We also describe application domains that can best exploit these types of hysteretic WTA circuits.

A two-dimensional object-based analog VLSI model of selective attentional processing has been implemented using a standard 1.2μm CMOS process. This chip extends previous work modeling object-based selection and scanning by incorporating the circuity and architectural changes necessary for two-dimensional focal plane processing. To balance the need for closely spaced, large photodetectors with the space requirements of complex in-pixel processing, the chip implements a multiresolution architecture. The system has the ability to group pixels into objects; this grouping is dynamic, driven solely by the segmentation criterion at the input. In the demonstration system, image intensity has been chosen for the input saliency map and the segmentation is based on spatial lowpass filtering followed by an intensity threshold. We present experimental results. 1

Analog very large-scale integrated (VLSI) circuits that perform the selection process for attentive sensory processing are presented. These circuits use excitatory feedback in a winner-take-all computation to produce a hysteresis in the selection from one location to the next. They also use an inhibitory mechanism to disengage attention so that more than one stimulus is attended, even when the inputs remain static. These circuits have been implemented in analog VLSI chips that sense and process optical input to select salient objects from the visual field. Data from the fabricated chips are shown, demonstrating

A system emulating the functionality of a moving eye---hence the name oculo-motor system---has been built and successfully tested. It is made of an optical device for shifting the field of view of an image sensor by up to 45 o in any direction, four neuromorphic analog VLSI circuits implementing an oculo-motor control loop, and some off-the-shelf electronics. The custom integrated circuits communicate with each other primarily by non-arbitrated address-event buses. The system implements the behaviors of saliency-based saccadic exploration,andsmooth pursuit of light spots. The duration of saccades ranges from 45 ms to 100 ms, which is comparable to human eye performance. Smooth pursuit operates on light sources moving at up to 50 o /s in the visual field. 1 INTRODUCTION Inspiration from biology has been recognized as a seminal approach to address some engineering challenges, particularly in the computational domain [1]. Researchers have borrowed architectures, operating ...