Supervised Learning Techniques for Detection of Underwater Acoustic Sources

Detection of acoustic sources in an ocean is an important and a challenging problem. In this work, we present a performance comparison of supervised learning techniques for detection of an acoustic source. Performance evaluation of the considered classifiers, which include support vector machine (SVM), logistic regression, decision trees, random forest and naive Bayes, are carried out on synthetic and experimental datasets. The noise model is assumed to follow a generalized Gaussian distribution with heavy tails, which depicts a practical scenario. Our extensive set of experiments indicate that SVM outperforms all other classifiers, in terms of detection accuracy.