Microsoft's artificial nose is able to recognize smells!

We are one step closer to human-like AI Robots.

With a simple gas sensor and a micro-controller, AI nose can identify the smell of bread, coffee, and many more scents.

The Artificial Nose uses a neural network to correlate the concentration of gases in the air to categories of smell.

When connected to an IoT platform, this nose could be used for a variety of scenarios: it could help to craft a real-time alert system for when foods have spoiled, or for detecting specific gases in the air.

Gas sensors measure the concentration of various gases emanating from a substance. An AI model extract the key characteristics of these gases and uses them to infer the corresponding smell.

The inferred result is shown to the user and can also be sent to Azure IoT to be further analyzed (ex. send alerts), or to contribute to updating and improving the model.

Building this Artificial Nose starts with the hardware, an off-the-self gas sensor and a microcontroller. The sensors can detect the concentration of gases—like Carbon monoxide (CO), Nitrogen dioxide (NO2), Ethyl alcohol(C2H5OH), and Volatile Organic Compounds (VOC)—found in the surrounding air.

The sensor is connected to a microcontroller, which is used to read and collect gas sensor data and feed it into the AI model.

The Artificial Nose “smells” by using a neural network to correlate the concentration of gases (CO, NO2, etc.) in the air—data inputs received from the gas sensor— to certain categories of smell (coffee, whiskey, bread).

The smell category is then displayed on the microcontroller screen along with a visual cue indicating the model’s degree of certainty in its correlation.