Rainfall intensity estimation from sound for generating CG of rainfall scenes

This paper suggests a method for creating computer graphics of rainfall scenes that naturally match the actual rainfall sounds. With respect to the rainfall phenomenon, Marshall and Palmer proposed probability distribution formulae of the number and size of raindrops in a certain volume under the given rainfall intensity. In addition, Best verified that the terminal velocity of a water droplet only depends on its size. These studies suggest us that we can derive the parameters of raindrops required for creating a rainfall animation, i.e. the distributions of the number and size of raindrops as well as their falling speed, only from the rainfall intensity. Hence, we propose an estimation method of the rain intensity from a real record of the rain sound. In the proposed method, the captured rain sound is Fourier transformed, and then two feature values are calculated in frequency domain. Finally, the rain intensity is estimated based on these feature values.