Odour Recognition with AI in Embedded Devices

OCEAN: Application for Optimising Energy Consumption in High-Power Electrical Devices
May 8, 2025
HARMONIZE: Enhanced Membrane Distillation and Crystallisation Technologies for Ammonia Product Recovery Approaching Zero Waste
May 13, 2025

12/05/2025

The digitalisation of human senses has advanced significantly in areas such as vision, hearing and touch. However, smell and taste continue to present a technological challenge, as their analysis relies on the identification of volatile or dissolved chemical compounds. While there are various sensors on the market capable of detecting specific gases and substances, they often lack the versatility required for comprehensive recognition of complex odours, showing limitations in adapting to different combinations of compounds.


In the field of artificial olfactory detection, various technologies have been explored that make use of the interaction between volatile compounds and sensing surfaces, generating changes in electrical properties such as conductivity. These phenomena—affected by factors like temperature and compound concentration—provide data that can be analysed to identify odours.

In this context, an innovative solution has been developed to offer greater flexibility and transparency in odour recognition using artificial intelligence. This approach focuses on the ability to easily and adaptively integrate new odours into the identification system. Furthermore, being an open-source solution, it gives users the freedom to customise and adapt it to their specific needs, fostering innovation and collaboration in the advancement of artificial olfaction.

The architecture of this solution is based on the use of efficient neural networks designed to run on low-cost microcontrollers. This enables the implementation of odour recognition systems in a wide range of devices and applications. A key feature of the solution is its ability to process sensory information in an optimised way, even allowing the sequential application of multiple AI models on the same acquired data. This feature opens the door to recognising a considerable number of odours using limited computational resources, enhancing the efficiency and versatility of olfactory detection systems.



Related Projects