Deterministic Multimodal Perturbation Enables Neuromorphic-Compatible Signal Multiplexing
Ui Jin Kim, Dong Hae Ho, Yoon Young Choi, Yongsuk Choi, Dong Gue Roe, Yonghyun Albert Kwon, Seongchan Kim, Young Jin Choi, Yejin Heo, Sae Byeok Jo, Geun Yeol Bae, Taeyoon Lee, Jeong Ho Cho
Abstract
Human multisensory neurons integrate multiple sensory information obtained from the external environment for precise interpretation of an event. Inspired by biological multisensory integration/multiplexing behavior, an artificial multimodal integration system capable of emulating the perception of discomfort based on the integration of multiple sensory signals is presented. The system utilizes a sensory ring oscillator that concisely and efficiently integrates thermosensory and hygrosensory signals from artificial receptors into voltage pulses whose amplitude and frequency reflect the two individual sensory signals. Subsequently, a synaptic transistor translates voltage pulses into a postsynaptic current, which exhibits a high correlation with the calculated humidex. Finally, the feasibility of the artificial multimodal integration system is successfully demonstrated using light-emitting diode discomfort indicators, suggesting that the proposed system can act as a foundation for future studies pertaining to neuromorphic perception and complex neurorobotics.