Underwater Acoustic Target Recognition based on Preemphasis Filter Convolutional Neural Network
Abstrak
N owadays, underwater acoustic target recognition (UATR) is a core technology in underwater acoustics. Identifying the underwater target from the complex marine background noise is critical to gain an acoustic advantage in underwater confrontation. To tackle the problem at its source, this paper proposes a deep learning model to improve target recognition capability based on the convolutional neural network (CNN) model with a preemphasis filter (PEF) module (herein presented as PEF+CN $N$ model). This model learns the fluctuations and differences between ship targets and multiple marine background noises. It then enhances the time-frequency spectrum quality and the signal-to-noise ratio (SNR) in a data-driven way. Finally, comparative experiments were conducted to validate the improvements of the proposed model. The results showed that the spectral feature is more pronounced than the original condition. Moreover, the overall and the single-class recognition accuracy also increased compared with the same CNN model without PEF, confirming the intelligence of the proposed model and offering the feasibility for more practical engineering applications.
Penulis (3)
Xiaopeng Kong
Yan Huang
Jingyi Wang
Akses Cepat
PDF tidak tersedia langsung
Cek di sumber asli →- Tahun Terbit
- 2024
- Bahasa
- en
- Sumber Database
- Semantic Scholar
- DOI
- 10.1109/COA58979.2024.10723512
- Akses
- Open Access ✓