Auto Detection of Blue Whale Calls
Abstract
The thesis establishes an auto detection system that obtains Nonnegative Matrix Factorization (NMF) and harmonic structure detection techniques and tests it on the variety of recording data which is collected by the vector sensor in Monterey Bay. This system is currently examined specifically on blue whale B calls. In the thesis, we find that a NMF can be implemented in different update rules and initialization methods and yield different output which depends on input data. In addition, we apply Genetic Algorithm (GA) to reduce the error which is produced by NMF. Last, Harmonic Product Spectrum (HPS) takes place to examine whether the harmonic structure of blue whale B calls exists in the output of NMF and GA. In order to reduce human resource on physically checking the existence of blue whale, the goal is to find the best combination of algorithms that identifies the calls with lowest rate of false alarm. In further research, the performance can be improved by applying parallel computation. And manipulating the data of vector sensor to locate the source of sound can increase the accuracy of this detection system. Moreover, obtaining the detection of various marine mammals is needed.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 2023
- Accession Number
- AD1224379
Entities
People
- Min-hsiu Wu
Organizations
- Naval Postgraduate School