Underwater Acoustic Localization with Bayesian Learning

About this movie

Underwater acoustic signals, or sound waves, are distorted by factors such as noise, refractions, and water depth, making it challenging to locate their source. The purpose of this project was to determine the efficacy of an adaptive learning algorithm in creating a model for estimating the location of an object emitting sound underwater. Using a hydrophone, we recorded subaquatic acoustic signals in the form of music to quantify the channel's distortions. We derived channel response coefficients using a Bayesian (or adaptive learning) model. With them, we calculated an absorption coefficient relating the distance between the sound and hydrophone to properties of the recording. Through the process of multilateration, we aimed to approximate, with minimal error, the location of the acoustic source. Our adaptive learning model converges on channel response coefficients with increasing accuracy, leaving a strong foundation for continued efforts in underwater acoustic localization. Produced and written by Miranda Schrade Faculty mentors: Dr. Yun Ye, Dr. Shenglan Yuan Department of Mathematics, Engineering, and Computer Science CUNY Research Scholar Program.

Customer Reviews