Poster Abstract: Liquid Identification via Container Acoustic Resonance

With the improvement in quality of life, ensuring liquid safety has become increasingly important, leading to a growing focus on liquid identification technologies. While extensive prior work has employed RF signals for this purpose, such methods typically require expensive and bulky signal transmitters or receivers. In this work, we explore a novel modality for liquid identification, i.e. container acoustic resonance. Our key observation is that the acoustic resonance spectrum of each liquid-container system is strongly correlated with the liquid's density and solute composition. We design and implement a simple prototype for liquid identification with low-cost acoustic sensors. The spectrum of channel frequency response for each liquid-container system is extracted as the criterion. The results show that the average accuracy for classifying 10 common liquids is 99.4% with ResNet-5.