Enabling Low-Power Signature Recognition for the IoT with SLIF Neurons

Energy constraints are still a significant challenge in numerous IoT applications, particularly due to the excessive power consumption of microcontrollers. To overcome this limitation, novel circuit designs have been introduced, with the integration of spiking neurons and analog computing emerging as a promising solution, facilitating substantial reductions in power consumption. However, the operation within the analog domain introduces complexities in managing the sequential processing of incoming signals, a critical requirement for diverse applications. This study employs the Saturating Synapses Leaky Integrate and Fire (SLIF) model, a bio-inspired neuron model, to develop a signature recognition system based on a Spiking Neural Network, without the need of non-biological techniques such as synaptic delays. SLIF neurons exhibit spiking behavior exclusively in response to two consecutive spikes with an Inter Spike Timing (IST) within a specific range, remaining unresponsive to other ISTs. We present the joint design of IST-based signatures and the corresponding network. Subsequently, we evaluate the system's efficiency in recognizing its specific sequence and discriminating against alternative sequences. The novelty of this paper lies in the proposition of a new type of temporal sequence recognition networks based on ISTs, offering significantly lower energy consumption compared to conventional approaches.