Precipitation Measurement from Raindrops’ Sound and Touch Signals

This study proposes a novel method for rainfall intensity estimation from acoustic and vibration data with low-cost sensors. At first, a precipitation measurement device was developed to collect sound and touch signals from raindrops, composed of Raspberry Pi, a condenser microphone, and an accelerometer with 6 degrees of freedom. To figure out whether rainfall occurred or not, a binary classification model with the XGBoost algorithm was considered to analyze long-term time series of vibration data. Then, high-resolution acoustic data was used to investigate the main characteristics of rainfall patterns at a frequency domain for the period when it was determined that rainfall occurred. As a result of the Short Time Fourier Transform (STFT), the highest frequency, mean and standard deviation of amplitudes were selected as representative values for minute data. Finally, different types of regression models were applied to develop the method for rainfall intensity estimation from comparative analysis with other precipitation measurement devices (e.g., PARSIVEL, etc.). It should be noted that the new device with the proposed method functions reliably under extreme environmental conditions when the estimated rainfall intensity was compared with measured data from ground-based precipitation devices. It shows that low-cost sensors with sound and touch signals from raindrops can be effectively used for rainfall intensity estimation with easy installation and maintenance, indicating a strong possibility of being considered in a wide range of areas for precipitation measurement with high resolution and accuracy

Acknowledgement

This work was supported by the National Research Foundation of Korea(NRF) grant funded by the Korea government(MSIT) (No. NRF-2022R1A4A3032838).