A Method for Constructing an Olfactory Dataset Based on fNIRS.

Biomimetic olfaction models have been applied to various fields such as image recognition, text classification, electroencephalography (EEG) recognition, and spatio-temporal sequence prediction. However, the challenge of obtaining olfactory data complicates the construction of olfactory datasets, ultimately resulting in a scarcity of applications for biomimetic olfaction models in olfactory tasks. In this work, we begin by conducting a comprehensive comparison of existing biomimetic olfactory models from various aspects such as structure, degree of biomimicry, and performance. Following this, prefrontal lobe olfactory data from ten subjects were collected for three different odors using functional Near-Infrared Spectroscopy (fNIRS). Finally, the data underwent denoising through the use of various data processing methods, and the quantity of the olfactory dataset was expanded. The olfactory dataset constructed in this study can be used to evaluate the performance of biomimetic olfactory models in processing olfactory tasks, which may aid in advancing the development of these models in real-world olfactory applications.