TL;DR
This paper introduces a novel method for detecting and differentiating low-frequency electric field signals from ships in complex underwater environments. By combining Empirical Mode Decomposition (EMD) with permutation entropy (PE) applied to the highest energy intrinsic mode functions (IMFs), the approach significantly improves the accuracy of target detection and reduces false alarms.
What Problem Does This Paper Address?
Detecting low-frequency electric field signals from ships underwater is challenging due to the complex marine environment, signal attenuation, and the non-stationary nature of these signals. Traditional methods struggle to extract reliable features, especially when subtle differences exist between vessel types, leading to false alarms. This paper aims to improve the accuracy and robustness of detecting and differentiating these signals by developing a more sophisticated feature extraction technique.
Methods
The core of the proposed method involves a two-pronged approach:
- Empirical Mode Decomposition (EMD): This technique decomposes the complex, non-stationary electric field signals into a series of Intrinsic Mode Functions (IMFs). EMD is chosen for its adaptive nature, breaking down signals based on their time-scale characteristics without requiring predefined basis functions, which is crucial for handling the non-stationary nature of the signals.
- Permutation Entropy (PE) on High-Energy IMFs: After EMD, the paper focuses on the IMFs that contain the most significant energy. Permutation entropy is then calculated for these IMFs. PE is a measure of the complexity and randomness of a time series. By analyzing the PE of the highest energy IMFs, the researchers aim to capture discriminative features that differentiate between various vessel types. The method specifically selects the IMF with the maximum average energy for feature extraction.
The paper also compares EMD with other decomposition methods like Empirical Wavelet Transform (EWT), Ensemble Empirical Mode Decomposition (EEMD), and Improved Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (ICCEMDAN) to demonstrate EMD's suitability for this task.
Datasets
- Simulated Data: The study utilizes a scaled model system to generate simulated shaft frequency electric field signals from three different types of vessels (Vessel A, B, and C). This system uses titanium-based oxide electrodes as dipole sources to mimic vessel emissions.
- Real Vessel Data: Offshore experiments were conducted near Zhifu Island, Yantai, using Ag/AgCl electrodes to collect electric field signals from three actual vessels (Vessel A, Vessel B, and Vessel C) with known parameters.
Key Insights and Ideas
- EMD is Superior for Shaft Frequency Signals: The study found that EMD, compared to EWT, EEMD, and ICCEMDAN, is more effective at extracting time-domain features of vessel shaft frequency electric fields and is better at handling noise.
- Energy-Based IMF Selection is Crucial: Focusing on the IMFs with the highest energy content after EMD is key. These components carry the most discriminative information for differentiating vessel types.
- Permutation Entropy Reveals Subtle Differences: While raw signal PE values showed limited differentiation, the PE of the highest energy IMFs revealed significant inter-vessel differences, making it a powerful feature for classification.
- Combined Approach Enhances Discrimination: The synergy between EMD's decomposition capability and PE's complexity analysis provides a robust method for distinguishing between vessels that might otherwise be indistinguishable due to minor variations in their electric field signatures.
Key Results
- The proposed method successfully screened different types of typical vessels.
- The entropy difference was improved from 0.05 to about 0.2.
- The difference rate of the shaft frequency electric field signal was improved by 75%.
- This led to a significant reduction in the false alarm rate for target detection.
- In real vessel data verification, the PE values of the IMF with maximum average energy showed a difference of approximately 0.5 between Vessel A and Vessel B, and 0.3 between Vessel A and Vessel C, indicating good differentiation.
Limitations and Future Work
- Vessel Type Recognition: While the method effectively distinguishes between different types of vessels, the paper acknowledges that recognizing specific vessel identities is not yet achieved.
- Future Work: The authors suggest that future research should focus on developing methods for vessel-type recognition based on the extracted shaft frequency electric field features.
Summary updated