Related Experiment Video
Updated: Jun 23, 2026
11:15
Applications of EEG Neuroimaging Data: Event-related Potentials, Spectral Power, and Multiscale Entropy
Published on: June 27, 2013
33.6K
Multiscale Sample Entropy-Based Feature Extraction with Gaussian Mixture Model for Detection and Classification of
Oluwaseyi Paul Babalola1, Olayinka Olaolu Ogundile2, Vipin Balyan1
1French-South African Institute of Technology, Department of Electrical, Electronic, and Computer Engineering, Cape Peninsula University of Technology, Bellville, Cape Town 7535, South Africa.
Entropy (Basel, Switzerland)
|April 26, 2025
Summary
A new multiscale sample entropy (MSE) algorithm improves blue whale call detection and classification. This method outperforms traditional techniques, offering higher accuracy for analyzing complex whale vocalizations.
Area of Science:
- Bioacoustics
- Signal Processing
- Machine Learning
Background:
- Continuous acoustic monitoring is crucial for understanding marine mammal vocal behavior.
- Analyzing complex and non-stationary vocalizations like those of blue whales presents significant challenges.
- Existing feature extraction methods may not fully capture the intricacies of blue whale calls.
Purpose of the Study:
- To introduce a novel multiscale sample entropy (MSE) algorithm for time-domain feature extraction in blue whale acoustics.
- To apply MSE in conjunction with Gaussian mixture models (GMM) for enhanced blue whale call detection and classification.
- To benchmark the proposed MSE-GMM algorithm against established methods.
Main Methods:
- Development and application of a multiscale sample entropy (MSE) algorithm for feature extraction.
- Integration of MSE features into a Gaussian mixture model (GMM) framework.
- Comparative analysis against Principal Component Analysis (PCA)-GMM, Wavelet-based Feature (WF)-GMM, and Dynamic Mode Decomposition (DMD)-GMM.
- Proposal of a GMM-based feature selection method considering inter-feature correlations.
Main Results:
- The proposed MSE-GMM algorithm demonstrated superior performance in classifying blue whale vocalizations.
- Achieved higher accuracy and lower error rates compared to PCA-GMM, DMD-GMM, and WF-GMM.
- The GMM-based feature selection enhanced classification model accuracy by evaluating correlated features.
Conclusions:
- Multiscale sample entropy (MSE) is an effective time-domain feature extraction technique for blue whale bioacoustics.
- The MSE-GMM approach offers a significant advancement for automated detection and classification of blue whale calls.
- This study provides a robust method for analyzing complex marine mammal vocalizations.
