Neuromorphic Auditory Sensor IPcore for Spiking Edge-AI
In NASSAI, the neuromorphic auditory sensor (NAS) FPGA implementation and its VLSI ASIC fabrication represent an improvement from previous TRL3 (proof of concept demos) to TRL5 (validation of the components in a relevant scenario). This novel way to process the sound wave using Spike-Signal-Processing (SSP) building-blocks represents the key innovative aspect of the NAS, taking low-power and low-latency benefits from neuromorphic systems because of its spike-based representation of the information. Our NAS transforms the information in the acoustic wave into an equivalent spike-rate representation through a simple digital circuit, called a spike-generator. Then, the NAS uses a set of cascade spike-based low-pass filters (SLPFs) bank inspired by Lyon’s biological cochlea model. The NAS has always been synthesised for a particular application, fixing in advance all the internal parameters (gain, cutoff frequencies, spiking output bandwidth, etc). In this PCD project we plan to move forward and offer a real product as an IPcore fully configurable after synthesis for any target application, study the market opportunities of the system distributed in a commercial MPSoC, like a Zynq UltraScale+ from Xilinx/AMD, and obtain a certification for validating it use in the automotive industry (voice commands) or medical sector (acoustic signal classification) in combination with Spiking-Neural-Networks. We aim to improve TRL from 6 to 7-8.
PIs: Alejandro Linares Barranco / Ángel Francisco Jiménez Fernándes
Type: Programa Estatal de I+D+i Retos de la Sociedad: Pruebas de Concepto
Reference: PDC2023-145841-C33
Funding by: Ministerio de Ciencia e Innovación
Start date: 01-01-2024
End date: 31-12-2025
Researchers:
Casanueva Morato, Daniel
Amaya Rodríguez, Claudio Antonio
Montes Sánchez, Juan Manuel
Gómez Rodríguez, Francisco de Asís
Vicente Díaz, Saturnino
Miró Amarante, Lourdes
Domínguez Morales, Juan Pedro
Ríos Navarro, José Antonio
Cerezuela Escudero, Elena
Piñero Fuentes, Enrique
Jiménez Moreno, Gabriel
Muñoz González, Tomás
More Information:
