An Optical Neuromorphic Sensor with High Uniformity and High Linearity for Indoor Visible Light Localization

• Published in: Advanced Sensor Research Vol. 3; no. 6
• Main Authors: Zhong, Shuai, Zhou, Jiachao, Yu, Fangwen, Xu, Mingkun, Zhang, Yishu, Yu, Bin, Zhao, Rong
• Format: Journal Article
• Language: English
• Published: Wiley-VCH 01.06.2024
• Subjects: illuminance encoding; indoor visible light localization system; optical neuromorphic sensor; threshold switching memristor
• ISSN: 2751-1219; 2751-1219
• DOI: 10.1002/adsr.202300197

The visible light localization system holds great promise as a highly accurate indoor positioning method. However, it still suffers deficiencies including high latency and power consumption, and large area cost. To address these issues, a high energy efficient spiking localization system inspired by the biological spatial representation system is presented. This system utilizes an optical neuromorphic sensor, consisting of a compact NbOx‐based threshold switching memristor and a photoresistor. The key lies in the system's ability to convert analog light information into electrical spikes, resembling the behavior of sensory neurons, which enables the encoding of light illuminance through spiking frequency. Consequently, the system achieves high uniformity, high linearity (≈10%), and high sensitivity (≈1.1 kHz Lux−1 and ≈72.7 kHz cm−1 for light illuminance and distance detection, respectively), indicating its potential suitability for visible light localizations. By leveraging a spiking neural network classifier, the system successfully distinguishes locations with different illuminances. After 150 epochs, it achieves an accuracy of 97%, showcasing the feasibility of using the spiking localization system in real‐world applications. The approach of spike‐based light positioning is a leap forward toward the development of future compact, highly energy‐efficient visible light localization systems. A high energy‐efficient, spike‐based indoor positioning system inspired by biological spatial representation is reported. Using an NbOx‐based optical neuromorphic sensor with high uniformity and linearity, positions with different illuminances can be encoded with spike frequency and an accuracy of 97% is achieved through spiking neural network classification, offering a promising solution for future compact, energy‐efficient indoor visible light localization systems.