Characterization and modeling of the sound production of beluga whales (Delphinapterus Leucas) based on models of analysis / speech synthesis




  Guillermo Fernán Lara Martínez


  Ramón Miralles Ricós


This thesis deals with the study of the sounds produced by beluga whales (Delphinapterus leucas) with a fundamental objective: characterization and modeling. To this end, a series of analysis / synthesis algorithms of the sounds produced by these animals inspired by recent researches on the functioning and the physiognomy of the vocal apparatus of beluga whales is proposed.

It is a multidisciplinary work in which to achieve this goal, the generation of sounds in musical instruments is analyzed, techniques time-frequency analysis are reviewed along with methods of pattern recognition, screening techniques features are applied, the potential to include algorithms that work in the cepstral domain is analyze, and a quantitative analysis of the Recurrence Plots is performed. All this allows to propose a model of sound production capable to adapt to the peculiarities of this species and reproduce with high fidelity its wide repertoire of sounds. In addition, in order to validate the proposed model, it will be analyze and synthesize different sounds of a database of beluga whale vocalizations to compare with those generated with a generic analysis / synthesis model.

Additionally, it is proposed to use the synthesis model parameters for a new sound classification based on their nature of production, achieving better results with classifiers based on characteristics of time-frequency diagram.The proposed hypotheses have been validated with performing acoustic measurements of beluga whales Oceanografic of Valencia (supervised by park biologist), as well as a large number of audio laboratory recordings under controlled conditions.

Finally it is explained the device of passive acoustic monitoring SAMARUC, designed with functionality to include algorithms implemented in an environment of open water and capable of providing the recorded sounds, their classification, as well as indicators of good environmental status of our seas and oceans. These indicators include underwater noise levels as reflected in the descriptor 11 of the European Marine Directive.