Applications of machine learning in induction cooking

In this paper we present a new line of work combining digital signal processing and machine learning algorithms to identify and classify cooking recipients used in domestic induction heating. First, a recipient `signature' is obtained from processing current and voltage waveforms, which includes impedance harmonics and the power factor. Then, the non-linear fitting capabilities of artificial neural networks and other machine learning algorithms allow processing the pot signature. We show two applications of this technique: recipient size estimation and identification of every cooking recipient of a specific user (for instance, for assigning a specific cooking profile to each vessel). Finally, we implement our procedure onto a low-cost electronic circuit such as those included in commercial induction home appliances. This new approach is of interest for developing new applications in the context of automatic cooking.