Neuromorphic analog spiking-modulator for audio signal processing
- PDF / 2,737,847 Bytes
- 16 Pages / 595.276 x 790.866 pts Page_size
- 58 Downloads / 227 Views
Neuromorphic analog spiking-modulator for audio signal processing Pietro M. Ferreira1,2
•
Jamel Nebhen3
•
Geoffroy Klisnick1,2 • Aziz Benlarbi-Delai1,2
Received: 14 April 2020 / Revised: 5 September 2020 / Accepted: 28 September 2020 Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract While CMOS scaling is currently reaching its limits in power dissipation and circuit density, the analogy between biology and silicon is emerging as a solution to ultra-low-power signal processing. Urgent applications involving artificial vision and audition, including intelligent sensing, appeal original energy efficient and ultra-miniaturized silicon-based solutions. While state-of-the-art is focusing on digital-oriented solutions, this paper proposes a neuromorphic analog signal processor using Izhikevich-based artificial neurons in an analog spiking modulator. A varicap-based artificial neuron is explored reducing the silicon area to 98:6 lm2 and the substrate leakage to a 1:95 fJ=spike efficiency. Post-layout simulation results are presented to investigate the high-resolution, high-speed, and full-scale dynamic range for audio signal processing applications. The proposal demonstrates a 9 bits spiking-modulator resolution, a maximum of 8 fJ=conv efficiency, and a root–mean–square error of 0:63 mVRMS . Keywords Artificial neuron Spiking signal processing Non-linear electronics Ultra-low power
1 Introduction Neuromorphic computing appeared in the 90 s as a complementary architecture to von Neumann systems using analog circuits designed to mimic biological neural systems [1]. Since then, digital neuromorphic systems have been often implemented in FPGAs considering its shorter design and manufacturing time, reconfigurability and reusability for different applications [2]. From both & Pietro M. Ferreira [email protected] Jamel Nebhen [email protected] Geoffroy Klisnick [email protected] Aziz Benlarbi-Delai [email protected] 1
Universite´ Paris-Saclay, CentraleSupe´lec, CNRS, Lab. de Ge´nie E´lectrique et E´lectronique de Paris, 91192 Gif-sur-Yvette, France
2
Sorbonne Universite´, CNRS, Lab. de Ge´nie E´lectrique et E´lectronique de Paris, 75252 Paris, France
3
Prince Sattam bin Abdulaziz University, College of Computer Engineering and Sciences, P.O. Box 151, Alkharj 11942, Saudi Arabia
circuitries, analog one has often been a good solution to implement the processing components of neurons and synapses, due to its ability to faithfully mimic biological systems. Besides, analog solutions have presented the best energy consumption per unit of information, often represented in J=spike. However, analog solutions have presented several reliability challenges in terms of process, temperature, and voltage variation that are frequently overcome by digital solutions [3]. A mixed circuit approach usually presents a better trade-off between digital and analog solutions [4]. Neuromorphic spiking signal processors are highly energy-efficient, parallel and di
Data Loading...
