Self-organising sensory maps in odour classification mimicking

Area of Science:

• Artificial intelligence
• Sensory systems engineering
• Biomimetic systems

Background:

• Traditional artificial olfaction systems face challenges in accurate odour classification and efficient hardware implementation.
• Self-Organising sensory Map (SOM)-based algorithms offer a promising approach for pattern recognition in sensory data.

Purpose of the Study:

• To introduce a novel artificial olfaction system integrating a gas sensor array with a self-organising artificial neural network.
• To reformulate and detail effective SOM-based algorithms for odour classification.
• To present an innovative opto-electronic micromachined implementation of the neural network.

Main Methods:

• Development of a gas sensor array for detecting volatile compounds.
• Application and reformulation of Self-Organising sensory Map (SOM) algorithms for odour data processing.
• Design and implementation of an opto-electronic micromachined neural network with a novel hybrid activation mechanism.

Main Results:

• Successful integration of a sensor array and a self-organising artificial neural network for artificial olfaction.
• Detailed reformulation of SOM algorithms enhancing odour classification capabilities.
• Introduction of a novel hardware implementation for the neural network, avoiding fabricated cloning templates.

Conclusions:

• The developed artificial olfaction system demonstrates enhanced odour classification capabilities.
• The novel opto-electronic micromachined implementation offers an efficient hardware solution for artificial neural networks in sensory applications.
• This work advances the field of artificial olfaction and biomimetic sensory systems.