2025_programme: A deep-sea ambient noise prediction method with few training samples based on VMD and LSTM

• Day: June 17, Tuesday
    Location / Time: C. THALIA at 10:10 - 10:30
• Last minutes changes: -
• Session: 23. Underwater noise modelling and measurements
  Organiser(s): David Barclay, Martin Siderius
  Chairperson(s): David Barclay, Martin Siderius
• Lecture: A deep-sea ambient noise prediction method with few training samples based on VMD and LSTM
  Paper ID: 2168
  Author(s): Guoli Song, Xinyi Guo, Qianchu Zhang, Bo Yuan, Qunyan Ren, Li Ma
  Presenter: Guoli Song
  Abstract: Deep-sea ambient noise levels are an important factor affecting underwater target detection, underwater acoustic communications, as well as marine ambient monitoring and animal protection. In order to predict the deep-sea ambient noise level, a prediction method is established combined with Variational Mode Decomposition (VMD) and Long Short-Term Memory Networks (LSTM). It can realize the ambient noise level prediction under the condition of few training samples in the frequency band of 20Hz-10kHz. The test data verifies that the level of the ambient noise in the next 7 days can be predicted by using the current 3 days deep-sea ambient noise data. And the frequency and time trends of the forecast results are consistent with the measured noise trends. The average of the root mean square error (RMSE) is less than 1 dB in both the prediction band and the prediction time.
    Download the full paper
  This paper is a candidate for the "Prof. John Papadakis award for the best paper presented by a young acoustician(under 40)"
• Corresponding author: Dr Guoli Song
  Affiliation: Institute of Acoustics, Chinese Academy of Sciences
  Country: China