This paper presents a biomimetic intelligent olfactory sensing system inspired by the *Drosophila* olfactory system. The system integrates an 18-channel MEMS sensor array (16 gas sensors, 1 humidity sensor, 1 temperature sensor), a CMOS circuit, and a lightweight machine-learning algorithm. The system processes olfactory data using a combination of a shallow neural network and a residual neural network to identify 7 types of gases with 98.5% accuracy and predict gas concentrations with 93.2% accuracy, even in challenging high-humidity or partially damaged sensor scenarios. This system is designed for applications such as emergency rescue alarm systems.

Xiawei Yue, Jiachuang Wang, Heng Yang, Zening Li, Fangyu Zhao, Wenyuan Liu, Pingping Zhang, Hong Chen, Hanjun Jiang, Nan Qin, Tiger H. Tao