An acoustic recognition system of the individual dolphin

Abstract

Nowadays, human pays much more attention to protect dolphins increasingly. Therefore, the aims of my project are to develop a recording unit that can sit into the sea in order to record the whistles of the dolphins and to monitor the dolphins. To illustrate in more precise way, my project focuses on building on an acoustic recognition system for individual dolphin. This is similar to build a system for speaker recognition for human speech. As a result, my approach is to build a system for speaker recognition first. There are two main modules for speaker recognition, namely, feature extraction and pattern matching. For feature extraction, Mel – Frequency Cepstrals Coefficients has been adopted while Vector quantization has been adopted in pattern matching. Besides, all the algorithms involved in this project are implemented by MATLAB. In this project, 6 speakers spoke ‘one ‘and ‘two’ respectively as the test cases for the recognition system which is to identify each of the speakers with any text constraints. As the results shown in this report, all speakers can be recognized successfully in text independent identification. After this step, the system can still recognize dolphins ‘whistles (4 dolphins in total) by applying the same methodology used in handling human voices.