An Oxide Based Spiking Thermoreceptor for Low-Power Thermography Edge Detection

Thermoreceptors can encode thermal stimuli into spikes that are processed by the neural network to endow human with thermal perception. Such energy-efficiency and robust interaction with real-world have inspired the rise of the artificial spiking thermoreceptor (AST). However, monolithic spiking thermoreceptor is still rarely reported, which may due to the lack of device that can simultaneously implement temperature-sensing and spike-encoding functions. Here, we demonstrate an artificial spiking thermoreceptor based on Ag/TaO
$_{X}$
/AlO
$_{X}$
/ITO threshold switching memristor to achieve human-like thermal perception. The device is able to encode thermal information into spikes at a low power consumption (< 240 nW). These advantages consequently facilitate the demonstration of power efficient and accurate thermography edge detection based on the array of such AST and a pulse coupled neural network (PCNN).