Low Frequency Underwater Acoustic Modelling Based on Deep Learning

Abstract

In ocean environments, acoustic communication is the most efficient method and the study of the behaviour of sound and its nature in the ocean is called ocean acoustics. Modelling of ocean acoustic propagation is an essential step for designing acoustic devices as the environmental behaviour of the ocean varies with climate, seasons, aquatic life and other form of chemical reactions. Modelling acoustic propagation based on parabolic equations is one of the most efficient ways, especially with low-frequency applications. The computational complexity and the time for modelling the parabolic equations are resolved by the predictive modelling presented in this paper. Modelling of several environments with different acoustic parameters generates the necessary data for the predictive modelling system. Deep learning is the process of learning data inspired by biological systems based on weights and biases. A custom deep learning model is developed for the understanding of transmission loss data of different ocean environments. In acoustic communication, transmission loss is the decrease in the sound signal by the ocean environmental parameters. A user application is developed that resembles the traditional modelling but is backed by a deep learning system that predicts the transmission loss for the entire environment.