Wake-up radio receiver based on Spiking Neurons for detecting activation sequence

Abstract

Energy consumption is a critical issue for the deployed nodes in the area of Internet of Things (IoT). This is the reason why many research projects focus on Wake-up Radio (WuR) receivers that permit the nodes to remain in sleep mode for as long as possible and to wake them up only if needed. However, current WuR use classic microcontrollers that are still too energy consuming. Meanwhile, Spiking Neural Networks (SNN) offer much lower power consumption. Thus, we propose to adapt those neural networks as a wake-up radio receiver in the IoT context. We aim at waking up the concerned node by recognising one or many activation sequences in a bit flow. We propose here a configuration for the neurons along with the design of appropriate sequences. We present the performances of our system and the impact of different parameters on the accuracy of the recognition system.