Development and Optimization of an Artificial Olfactory Bulb

This study combines thermal modulation technology with a metal oxide (MOX) sensor array to construct an artificial olfactory bulb (OB) comprising 450 response units (olfactory glomeruli). Compared to the unmodulated control group, this artificial OB both exhibits significant performance improvement and mimics the mammalian olfactory system by presenting odor information in a spatially encoded manner. Inspired by the weight distribution mechanism in the mammalian OB, a wavelet-based two-dimensional decomposition and reconstruction was applied for weight assignment to the designated core glomeruli (90 glomeruli) and secondary glomeruli (360 glomeruli), which were selected by a novel contribution analysis algorithm based on silhouette. Experimental results indicate that after wavelet reconstruction, the efficiency of spatial encoding is enhanced, with the average qualitative accuracy of six classifiers increasing from 89% to 96%. These findings demonstrate that this bio-inspired approach can effectively enhance the performance of artificial olfactory bulb, providing insight for high-precision odor recognition.