Odor Recognition with a Spiking Neural Network for Bioelectronic Nose

Area of Science:

• Neuroscience
• Bioelectronics
• Artificial Intelligence

Background:

• Electronic noses struggle with complex odors, unlike biological systems.
• Mammalian olfactory bulbs process odors via neural signals.
• Bioelectronic noses aim to mimic biological olfaction.

Purpose of the Study:

• To develop a spiking neural network (SNN)-based method for odor recognition using neural signals.
• To improve odor recognition accuracy by exploiting precise spike timing.
• To address overfitting in SNNs with a novel voltage-based regulation strategy.

Main Methods:

• Recording neural spike trains from the rat olfactory bulb using electrode arrays.
• Implementing a Spiking Neural Network (SNN) model to decode neural signals.
• Developing and applying a voltage-based regulation strategy for SNN training.

Main Results:

• The SNN-based odor recognition method achieved state-of-the-art performance.
• The proposed voltage regulation strategy improved SNN performance by approximately 15% compared to classical SNN models.
• The method effectively utilizes the timing information present in neural spike trains.

Conclusions:

• The developed SNN approach offers a promising method for bioelectronic nose odor recognition.
• The voltage-based regulation strategy enhances the robustness and accuracy of SNNs in olfactory applications.
• This research advances the integration of neuroscience and AI for improved sensory systems.