Estimation of harmonic signatures in underwater noise from small boats

Abstract

This thesis explores underwater noise from small boats. There has been an in- creased focus on anthropogenic noise in the marine environment within the EU in the past years. The focus of the thesis is to study the noise from the engines and propellers of small boats, which is in general composed of both a harmonic series, and broadband noise.

In this thesis, we apply two fundamental frequency estimators to a dataset con- sisting of noise recorded in the waters outside Vealøs in order to extract the harmonic signature of small vessels. We will explore estimators based upon the Non-linear Least Squares method and Capons method, which can calculate the fundamental frequency of harmonic noise in water. The estimators are adapted in order to suit the dataset and evaluated based on accuracy and computation time. Results show that the Capon estimator excels in performance at lower signal to noise ratio with a slight advantage, but that it comes at a cost; the Non-Linear Least Squares estimator has approximately one sixth of the Capon estimators computation time. Since the estimator based on Non-linear Least Squares method has so much lower computation time, this estimator is certainly favourable as this fact opens possibilities for an increased number of samples with lower or comparable computation time.