Abstract
An MCU(microcontroller unit)-based edge computing system for assessing the height of liquids in containers in this paper. This system uses a solenoid valve to strike the bottle and a microphone to capture sound waves. Signals are transformed from the time domain to the frequency domain by FFT (Fast Fourier Transform) and used to predict the water level by using AI (artificial intelligence) model. A dynamic label smoothing method enhances label correlation, and an ANN (Artificial Neural Network) model is employed on the MCU for classification. The system accurately predicts water levels from 200 to 250 milliliters at 1-milliliter intervals. Hyperparameter optimization balances accuracy with MCU memory and computational constraints. Experimental results show that the system achieves an accuracy of 81% under the limits of edge computing, verifying its effectiveness in liquid level measurement applications.