LiquidListener: Supporting Ubiquitous Liquid Volume Sensing via Singing Sounds

This work proposes LiquidListener, a novel liquid volume sensing method for containers. Specifically, it enables the ubiquitous measurement of liquid volume not available in existing work due to
${i}$
) dependencies on dedicated sensing hardware (e.g., capacity sensors) and containers (e.g., transparent containers) and ii) a high training intensity. A key enabler of LiquidListener is listening to singing sounds. When a user taps a container using solid objects, such as pens and teaspoons, the container vibrates freely and produces a singing sound. As the container is filled with more liquid, the pitch of the sound decreases. Based on this relationship, we develop acoustic-based liquid volume sensing algorithms that support the precise measurement of liquid volume while using only a smartphone and requiring minimal user effort for calibration. The extensive experiments demonstrate that LiquidListener can support high accuracy with an average error ratio of 2.3% in sensing the liquid volume in various containers. In addition, the experimental results indicate that it can still maintain a similar level of accuracy in diverse and dynamically changing environments, even without additional calibration.