Ralph Etienne-Cummings

Neuromorphic silicon neuron circuits

Hardware implementations of spiking neurons can be extremely useful for a large variety of applications, ranging from high-speed modeling of large-scale neural systems to real-time behaving systems, to bidirectional brain–machine interfaces. The specific circuit solutions used to implement silicon neurons depend on the application requirements. In this paper we describe the most common building blocks and techniques used to implement these circuits, and present an overview of a wide range of neuromorphic silicon neurons, which implement different computational models, ranging from biophysically realistic and conductance-based Hodgkin–Huxley models to bi-dimensional generalized adaptive integrate and fire models. We compare the different design methodologies used for each silicon neuron design described, and demonstrate their features with experimental results, measured from a wide range of fabricated VLSI chips.

A biomorphic digital image sensor

An arbitrated address-event imager has been designed and fabricated in a 0.6-/spl mu/m CMOS process. The imager is composed of 80 /spl times/ 60 pixels of 32 /spl times/ 30 /spl mu/m. The value of the light intensity collected by each photosensitive element is inversely proportional to the pixels interspike time interval. The readout of each spike is initiated by the individual pixel; therefore, the available output bandwidth is allocated according to pixel output demand. This encoding of light intensities favors brighter pixels, equalizes the number of integrated photons across light intensity, and minimizes power consumption. Tests conducted on the imager showed a large output dynamic range of 180 dB (under bright local illumination) for an individual pixel. The array, on the other hand, produced a dynamic range of 120 dB (under uniform bright illumination and when no lower bound was placed on the update rate per pixel). The dynamic range is 48.9 dB value at 30-pixel updates/s. Power consumption is 3.4 mW in uniform indoor light and a mean event rate of 200 kHz, which updates each pixel 41.6 times per second. The imager is capable of updating each pixel 8.3K times per second (under bright local illumination).

Decoding of Individuated Finger Movements Using Surface Electromyography

Upper limb prostheses are increasingly resembling the limbs they seek to replace in both form and functionality, including the design and development of multifingered hands and wrists. Hence, it becomes necessary to control large numbers of degrees of freedom (DOFs), required for individuated finger movements, preferably using noninvasive signals. While existing control paradigms are typically used to drive a single-DOF hook-based configurations, dexterous tasks such as individual finger movements would require more elaborate control schemes. We show that it is possible to decode individual flexion and extension movements of each finger (ten movements) with greater than 90% accuracy in a transradial amputee using only noninvasive surface myoelectric signals. Further, comparison of decoding accuracy from a transradial amputee and able-bodied subjects shows no statistically significant difference ( p < 0.05) between these subjects. These results are encouraging for the development of real-time control strategies based on the surface myoelectric signal to control dexterous prosthetic hands.

Towards the Control of Individual Fingers of a Prosthetic Hand Using Surface EMG Signals

The fast pace of development of upper-limb prostheses requires a paradigm shift in EMG-based controls. Traditional control schemes are only capable of providing 2 degrees of freedom, which is insufficient for dexterous control of individual fingers. We present a framework where myoelectric signals from natural hand and finger movements can be decoded with a high accuracy. 32 surface-EMG electrodes were placed on the forearm of an able-bodied subject while performing individual finger movements. Using time-domain feature extraction methods as inputs to a neural network classifier, we show that 12 individuated flexion and extension movements of the fingers can be decoded with an accuracy higher than 98%. To our knowledge, this is the first instance in which such movements have been successfully decoded using surface-EMG. These preliminary findings provide a framework that will allow the results to be extended to non-invasive control of the next generation of upper-limb prostheses for amputees.

A foveated silicon retina for two-dimensional tracking

A silicon retina chip with a central foveal region for smooth-pursuit tracking and a peripheral region for saccadic target acquisition is presented. The foveal region contains a 9/spl times/9 dense array of large dynamic range photoreceptors and edge detectors. Two-dimensional direction of foveal motion is computed outside the imaging array. The peripheral region contains a sparse array of 19/spl times/17 similar, but larger, photoreceptors with in-pixel edge and temporal on-set detection. The coordinates of moving or flashing targets are computed with two one-dimensional centroid localization circuits located on the outskirts of the peripheral region. The chip is operational for ambient intensities ranging over six orders of magnitude, targets contrast as low as 10%, foveal speed ranging from 1.5 to 10 K pixels/s, and peripheral ON-set frequencies from <0.1 to 800 kHz. The chip is implemented in a 2 /spl mu/m n-well CMOS process and consumes 15 mW (Vdd=4 V) in normal indoor light (25 /spl mu/W/cm/sup 2/). It has been used as a person tracker in a smart surveillance system and a road follower in an autonomous navigation system.

An in silico central pattern generator: silicon oscillator, coupling, entrainment, and physical computation

Abstract. In biological systems, the task of computing a gait trajectory is shared between the biomechanical and nervous systems. We take the perspective that both of these seemingly different computations are examples of physical computation. Here we describe the progress that has been made toward building a minimal biped system that illustrates this idea. We embed a significant portion of the computation in physical devices, such as capacitors and transistors, to underline the potential power of emphasizing the understanding of physical computation. We describe results in the exploitation of physical computation by (1) using a passive knee to assist in dynamics computation, (2) using an oscillator to drive a monoped mechanism based on the passive knee, (3) using sensory entrainment to coordinate the mechanics with the neural oscillator, (4) coupling two such systems together mechanically at the hip and computationally via the resulting two oscillators to create a biped mechanism, and (5) demonstrating the resulting gait generation in the biped mechanism.

Asynchronous Decoding of Dexterous Finger Movements Using M1 Neurons

Previous efforts in brain-machine interfaces (BMI) have looked at decoding movement intent or hand and arm trajectory, but current cortical control strategies have not focused on the decoding of 3 actions such as finger movements. The present work demonstrates the asynchronous decoding (i.e., where cues indicating the onset of movement are not known) of individual and combined finger movements. Single-unit activities were recorded sequentially from a population of neurons in the M1 hand area of trained rhesus monkeys during flexion and extension movements of each finger and the wrist. Nonlinear filters were designed to detect the onset of movement and decode the movement type from randomly selected neuronal ensembles (assembled from individually recorded single-unit activities). Average asynchronous decoding accuracies as high as 99.8%, 96.2%, and 90.5%, were achieved for individuated finger and wrist movements with three monkeys. Average decoding accuracy was still 92.5% when combined movements of two fingers were included. These results demonstrate that it is possible to asynchronously decode dexterous finger movements from a neuronal ensemble with high accuracy. This work takes an important step towards the development of a BMI for direct neural control of a state-of-the-art, multifingered hand prosthesis.

Implementation of steerable spatiotemporal image filters on the focal plane

This paper presents an architectural overview of a pseudogeneral image processor (GIP) chip for realizing steerable spatial and temporal filters at the focal-plane. The convolution of the image with programmable kernels is realized with area-efficient and real-time circuits. The chips architecture allows photoreceptor cells to be small and densely packed by performing all analog computations on the read-out, away from the array. The size, configuration, and coefficients of the kernels can be varied on the fly. In addition to the raw intensity image, the chip outputs four processed images in parallel. The convolution is implemented with a digitally programmable analog processor, resulting in very low-power consumption at high-computation rates. A 16/spl times/16 pixels prototype of the GIP has been fabricated in a standard 1.2-/spl mu/m CMOS process and its spatiotemporal capabilities have been successfully tested. The chip exhibits 1 GOPS/mW at 20 kft/s while computing four spatiotemporal convolutions in parallel.

A Multichip Neuromorphic System for Spike-Based Visual Information Processing

We present a multichip, mixed-signal VLSI system for spike-based vision processing. The system consists of an 80 60 pixel neuromorphic retina and a 4800 neuron silicon cortex with 4,194,304 synapses. Its functionality is illustrated with experimental data on multiple components of an attention-based hierarchical model of cortical object recognition, including feature coding, salience detection, and foveation. This model exploits arbitrary and reconfigurable connectivity between cells in the multichip architecture, achieved by asynchronously routing neural spike events within and between chips according to a memory-based look-up table. Synaptic parameters, including conductance and reversal potential, are also stored in memory and are used to dynamically configure synapse circuits within the silicon neurons.

Continuous decoding of finger position from surface EMG signals for the control of powered prostheses

As development toward multi-fingered dexterous prosthetic hands continues, there is a growing need for more flexible and intuitive control schemes. Through the use of generalized electrode placement and well-established methods of pattern recognition, we have developed a basis for asynchronous decoding of finger positions. With the present method, correlations as large as 0.91 and mean overall decoding errors of ∼11% have been achieved with average decoding errors of between decoded and actual conformation of the metacarpophalangeal joints of individual fingers. It is hoped that these results will serve as a foundation from which to encourage further investigation into more intuitive methods of myoelectric control of powered upper limb prostheses.