Portable AI-powered spice recognition system using an eNose based on metal oxide gas sensors

In our daily lives, we use spices and herbs. There are thousands of different sorts of spices that surround words. And occasionally it’s difficult to distinguish between them. Furthermore, without specialized knowledge it is impossible to determine whether they are fresh or not. A challenging algorithm and highly sensitive sensors are needed to predict the labels and freshness of spices and herbs based primarily on their smell. In this paper, we present AI-powered spice recognition system (AISRS), which is made up of an array of 8 inexpensive BME688 digital tiny sensors are exploited to classify four different types of herbs and spices: clove, cinnamon, anise, and chamomile. The proposed eNose measures temperature, humidity, pressure, and gas concentrations for various types of spices and condiments. For every sort of class, we keep track of more than 10,000 readings. Through the use of assessment indexes at each level, we were able to determine whether or not algorithms such as k-NN, Random Forest, SVM, MLP, DT, and AdaBoost were successful. The Random Forest instantaneous classification algorithm performed the best among others where the success rate for predicting and differentiating between the four classes was better than 97 percent according to the validation data. These validation findings plus the eNose’s low power consumption (0.05 W) make it possible for it to be improved and used in portable and battery-operated applications in the future.