2025_programme: Spatial-frequency sparse adaptive learning for enhancing tonals in impulse noise
- Day: June 17, Tuesday
Location / Time: D. CHLOE at 15:10-15:30
- Last minutes changes: -
- Session: 06. Enhancing underwater acoustic sensing through machine learning
Organiser(s): Cheng Chi
Chairperson(s): Peng Xiao
- Lecture: Spatial-frequency sparse adaptive learning for enhancing tonals in impulse noise
Paper ID: 2320
Author(s): Fengdan Jiang, Cheng Chi, Chonglei Liu, Guanqun Wang, Shenglong Jin, Yu Li, Haining Huang
Presenter: Fengdan Jiang
Abstract: Enhancing and extracting sinusoidal signals (tonals) is an important topic in underwater acoustic signal processing. Unfortunately, the classical methods of enhancing tonals deteriorate in the background of high impulse noise. In this work, to mitigate the influence of impulse noise, we propose a spactial-frequency sparse adaptive learning method of enhancing tonals. Based on the tonal sparsity in the both spatial and frequency domain, spatial-frequency sparse beamforming can be considered to enhance tonals. However, due to the impact of impulse noise, the azimuth-frequency spectrum obtained contains numerous noise-induced artifacts, which lead to serious false alarms. The stationarity of tonals and the non-stationarity of impulse noise are utilized to overcome the shortcoming of spatial-frequence sparse beamforming. By employing simple adaptive learning scheme, we filter out non-stationary noise artifacts, thereby achieving a high output signal-to-noise ratio (SNR) for the tonals. The proposed method is validated both through simulated and experimental data. Compared with the reference methods including conventional beamforming (CBF), space-frequency sparse beamforming, and adaptive line enhancer (ALE), the proposed method exhibits superior azimuth resolution and processing gain, significantly improving detection performance in impulse noise.
- Corresponding author: Ms Fengdan Jiang
Affiliation: Institute of Acoustics, Chinese Academy of Sciences
Country: China
