When Machine Learning Boosts Sensing Performance: A Compact and Contactless Soil Moisture Sensor Example

Nowadays, even if the “machine learning” term has degenerated into a buzz-word, it encloses a large framework of techniques that has recently demonstrated a disruptive change of the performance in a large number of application, thanks to the increased amount of data and novel approaches to treat them. Some parts of these framework (such as multivariate analysis) have been known since a long time, however they have been rediscovered and improved thanks to this new application trend. In this paper we will show how the extraction of the information in sensing could be dramatically improved on the basis of multivariate analysis techniques. We demonstrate the capabilities in a novel compact implementation of a soil moisture sensor based on a contactless microwave impedance spectroscopy technique.