Recognition of Particle Impacts in Acoustic Fixing of Dust Flow Using an Artificial Neural Network

Abstract

The search for methods to capture and record data on the intensity of the saltation flux of large particles (bouncing over the surface) is an important task, since their movement causes the generation of dust aerosol in arid areas. The proposed approach simulates a prototype device capable of tracing the number of particles, in particular in dust storm conditions. The article proposes a method for analyzing audio recordings to detect particle impacts using a neural network approach. The spectrogram of the sound signal is analyzed. The neural network performs the recognition of the intensity and frequency of particle impacts. The accuracy of the neural network on a test sample obtained in natural conditions is 87.27%.