Visualization, detection and classification of Risso's and Pacific white-sided dolphins using an empirical mode decomposition-based process
| Autor: | Nicholas J. Kirsch, Mahdi H. Al-Badrawi, Kerri D. Seger, Yue Liang |
|---|---|
| Rok vydání: | 2020 |
| Předmět: |
Acoustics and Ultrasonics
Computer science business.industry Process (computing) Pattern recognition Mooring Population density Hilbert–Huang transform Visualization Marine mammal Arts and Humanities (miscellaneous) Robustness (computer science) Spectrogram Instrumentation (computer programming) Artificial intelligence Cluster analysis business |
| Zdroj: | The Journal of the Acoustical Society of America. 148:2766-2766 |
| ISSN: | 0001-4966 |
| DOI: | 10.1121/1.5147700 |
| Popis: | Risso's and Pacific white-sided dolphins occupy similar habitats and have been acoustically documented in northward habitat expansion above the Aleutian archipelago. Therefore, being able to differentiate between the two, either more easily during manual analysis or through automation, is important to track their distribution and to estimate their population densities. In our previous work, a semi-blind detection and classification (SDC) process based on Empirical Mode Decomposition (EMD) showed promise for detecting and clustering vocalizations and mooring noises from several sources in the Bering Sea. In this work, we incorporated other decomposition methods into this process, such as Variational Mode Decomposition (VMD), to visualize peak and notch patterns in pulsed signals. The proposed process is being tested on four datasets that vary by location, water depth, and recording instrumentation to quantify its robustness and identify and overcome its limitations. A case study to determine the usability of EMD/ VMD spectrograms in marine mammal click and buzz classification was performed by comparing peak and notch patterns of these two dolphin species. We will present the pros and cons of incorporating VMD spectrograms into the previously published EMD SDC process. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|