Pennsylvania State University’s innovation replicates human taste recognition.
Saptarshi Das and his doctoral student Harikrishnan Ravichandran from the Pennsylvania State University have developed a novel electronic tongue for monitoring the freshness of food. The hit rate should be 95 percent. The high accuracy also results from the use of artificial intelligence (AI) in the evaluation algorithm. Suspicious substance combinations are therefore better recognized.
Also for diagnostics
In addition to food safety, the smart electronic tongue can also be used in medical diagnostics. The sensor uses its AI to precisely detect various substances and classifies their quality, authenticity and freshness. This quickly makes it clear whether, for example, milk has been adulterated with water.
The human tongue consists of taste sensors that send their impressions to the gustatory cortex, a biological neural network. This area of the brain interprets the signals from the sensors. This goes beyond what is perceived by the taste receptors, which primarily recognize the five major categories of sweet, sour, bitter, salty and umami.
Works like brain
The experts first recreated the human tongue using electronic taste sensors. To mimic the gustatory cortex, the researchers also developed a neural network, a machine learning algorithm that mimics the brain in evaluating and understanding the data that the taste sensors collect.
“We had previously studied how the brain reacts to different tastes and mimicked this process by integrating various 2D materials to create a kind of blueprint for how AI can process sensory information like a human,” explains Ravichandran, explaining the approach .
Graphene recognizes taste
The artificial tongue consists of a graphene-based field effect transistor and a conductive device that can detect ions in the samples and are connected to an artificial neural network, which in turn has been trained on different data sets. What is crucial is that the sensors are not functionalized. One sensor detects different types of chemicals instead of having to use a sensor for each chemical.
The researchers imbued the neural network with 20 specific parameters, all related to how a liquid sample interacts with the sensor’s electrical properties. The AI was able to recognize samples based on these parameters set by the researchers. This means the developers achieve a hit rate of around 80 percent. When they switched on the neural network, the tongue reached 95 percent. (pressetext.com)
