Advances and Challenges in Deep Learning-driven Machine Olfaction Technology

Abstract

Machine Olfaction is a biomimetic detection technology that simulates biological olfaction, which integrates chemical sensor arrays
with pattern recognition algorithms to detect and identify odor molecules, and has prominent application value. The rise of deep learning has
broken through the limitations of traditional algorithms, significantly improving the recognition accuracy, generalization ability and environmental adaptability of Machine Olfaction Systems. This paper expounds this technology, elaborates on the core components of Machine
Olfaction and its compatibility with deep learning, analyzes the application details of mainstream models, discuss the application scenarios,
current bottlenecks and development trends, and offer reference for subsequent research. Deep learning can automatically extract the deep features of odors, address the shortcomings of traditional algorithms and facilitate the implementation of the technology. However, there are still
deficiencies in data, model lightweighting, and other aspects.