Fig 5 - uploaded by Thomas Maier
Content may be subject to copyright.
Source publication
This paper presents a 64 x 64 pixel temporal contrast vision sensor for the 8-15 mum thermal infrared spectral range. The device combines microbolometer detector technology with biology-inspired ('neuromorphic') focal-plane array (FPA) processing circuitry to implement an asynchronous, 'spiking' sensor array. The sensor's individual pixels operate...
Contexts in source publication
Context 1
... point after "reset" event) at the node V Diff . If V Diff exceeds or undershoots adjustable thresholds, marking an "event" that is detected by one of two voltage comparators in the event generator, the "reset signal" is briefly activated, whereby the "reset switch" is closed and the operating point of the amplifier is reset (compare also Fig. 5.). From these reset pulses, handshake signals LG22 .15 0 (Req) are derived, which control the transmission of a data packet containing the address of the respective active pixel address (x,y-address in the pixel field), via an asynchronous bus arbiter (AER 5 ). In this way temporal changes of object temperature are detected, encoded in ...
Context 2
... handshake signals LG22 .15 0 (Req) are derived, which control the transmission of a data packet containing the address of the respective active pixel address (x,y-address in the pixel field), via an asynchronous bus arbiter (AER 5 ). In this way temporal changes of object temperature are detected, encoded in the time domain, and communicated. Fig. 5 shows exemplary signal waveforms of bolometer temperature and V Diff over an arbitrary period of time. The rate of change is encoded in the inter-event intervals while the polarity of the gradient, going hotter or cooler, is determined by the signal reaching an upper or lower threshold respectively ("+" or "-"events). The polarity bit ...
Citations
... The theoretical scope of application for neuromorphic systems is large but their realworld use cases are currently limited by the lack of sensors able to communicate directly with bio-inspired neural networks which are event-driven and thus communicate through spikes. Only a few sensors natively output spikes and they are often dedicated and optimized for a particular type of signal, e.g., retinomorphic sensors for event-based vision processing [9], [10]. The scarcity of event-based sensors leads developers to use regular data output by conventional sensors and to add dedicated encoding layers to generate spikes from frame-based data. ...
Spiking Neural Networks (SNNs) are promising candidates for low-power and low-latency embedded artificial intelligence. However, those networks require event-based data produced by neuromorphic sensors which are not widely available, except for a few specialized devices like neuromorphic retinas. For other data types, a solution lies in the use of conventional sensors in conjunction with encoding layers. However, when performed in software, this solution can be detrimental to energy consumption or latency. Here we introduce a flexible design methodology for efficiently implementing, optimizing, and evaluating digital architectures of spike encoding integrating algorithms available in the literature. In order to quickly evaluate different hardware architectures and to tailor the solution to the application needs, our approach relies on High-Level Synthesis (HLS) tools and Python scripting. We illustrate the methodology by generating various digital architectures of two encoding algorithms taken from the literature and we evaluate their energy consumption and timing performances on Field Programmable Gate Arrays. This work could overcome the lack of neuromorphic sensors and accelerate the development of lower-power hardware SNNs.
