Scientists have built graphene-based devices that can be enjoyed with human accuracy with a breakthrough that brings artificial sensing closer to human capabilities.
The system uses machine learning to interpret chemical signals and classify flavors.
What sets the invention apart is the ability of an artificial taste system to work in the first wet state.
This feature allows you to better simulate how actual taste buds work in the human mouth.
Detects taste in real time
The device is constructed from layered graphene oxide within a nanofluid structure. Unlike previous attempts, both sensing and computing are combined into a single platform to more integrate the system than traditional artificial tongues.
Graphene oxides, such as pure graphene, react electrically when exposed to different chemicals.
The team trained the sensors using signals of 160 chemicals associated with a typical flavor profile.
These signals were fed to machine learning algorithms that construct a memory for how each flavor changes the conductivity of the material.
Our learning approach closely reflects how our brains communicate signals from taste bud. For a long time, it was believed that humans recognize five flavors: sweet, salty, bitter, sour, and Tamura. In 2023, scientists added a sixth ammonium chloride.
The artificial tongue focused on the first four. We have identified previously learned tastes with an accuracy of approximately 98.5%. When introduced into 40 new flavors, its accuracy ranged from 75% to 90%.
Researchers also taught them to recognize more complex combinations, including those found in coffee and cola.
credit – Proceedings of the National Academy of Sciences.
Graphene electric edge
Pure graphene was first separated in 2004 by Andre Geim and Kosta Novoslov. Its unique single-layer carbon atom lattice structure offers excellent electrical, mechanical and chemical properties.
The new sensors utilize the sensitivity of graphene oxide to chemical changes. When exposed to flavor compounds, it detects slight variations in conductivity and is extremely effective in pattern recognition when combined with machine learning.
“This system could one day restore taste to those who have lost their ability,” the author noted. They added that loss of taste can be attributed to stroke, viral infection, or degenerative disease.
For real-world applications
This innovation addresses the major limitations of previous artificial tasting systems, the separation of sensing and processing. The unified structure of the current model allows for faster and more efficient interpretation of taste data.
However, the device remains a proof of concept. It's still too bulky and is energy-intensive for consumer and medical use. Researchers say the next steps include miniaturization and power optimization.
If successful, the sensors can find use in food safety, quality control, and even robotics, if intelligent taste is required, beyond healthcare.
This study is published in the journal Proceedings of the National Academy of Sciences.
