Thesis Details

Passive acoustics is a very important field in so many areas. Although this field has been studied for a long time, scientists are already working on new techniques for detection, classification and localization, in order to improve the current algorithms; improving the accuracy, energy used, etc. The usefulness of this field depends on the area, but it can be implemented from marine animal research, anthropogenic levels monitoring, oil and gas facilities monitoring to military purposes.

Due to the little research made on the country, the need arises of make a model which is able to classify underwater acoustic signals based on its acoustic signature.

Considering some external factors such as ambient noise or others which could be present during the recording process. This will lead to develop an original recognition model.

A classifier based on ensemble methodologies was designed, having MLP neural networks as base classifiers and being able to classify up to 14 classes, whose signals go from 0 to 22050 Hz. The sounds involved were analyzed and the best feature extraction technique was chosen.

Good results were obtained in the classification of many underwater acoustic signals, considering the state of art, even proving that the architecture proposed can be used with more classes. These results serve as a basis to help monitoring México seas or any other place of interest, either real-time or deferred.

Honorable Mention - Given by the academic comitee evaluating the thesis.