2019_programme: APPLICATION OF GENERAL REGRESSION NEURAL NETWORK (GRNN) FOR UNDERWATER NOISE ESTIMATION

• Session: 27. Neural networks and info processing
  Organiser(s): N/A
• Lecture: APPLICATION OF GENERAL REGRESSION NEURAL NETWORK (GRNN) FOR UNDERWATER NOISE ESTIMATION
  Paper ID: 789
  Author(s): Wiem Belhedi, Francois Rioult, Achraf Drira, Medjber Bouzidi
  Presenter: Wiem Belhedi
  Presentation type: poster
  Abstract: With growing exploration and utilization of the ocean by human beings, sound pollution increases significantly. Hence, the impact of noise on marine organisms has become one of the most important research topics.\nHowever, simulating sound pollution starting from boat locations requires long computations. It is in this context that the proposed work is situated. In fact, the aim of this project is to study how deep learning techniques can speed up these long computations.\nThe proposed approach consists of General regression neural network (GRNN) in order to estimate the noise level starting from boat locations depending on several inputs. GRNN, which is a type of supervised network, is able to produce continuous value outputs and to approximate functions hence to perform a regression. This makes this model one of the suitable solutions for our task. The proposed GRNN architecture is a three-layer (input, hidden and output layer) network where there is one hidden neuron for each training pattern in hidden layer.\nExperimental data from practical investigations/ experiments were used to train the GRNN for estimating noise level caused by each boat. The predicted values using GRNN closely followed the experimental ones with an average mean squared error scores (RMSE) less than 12.9%. Results show that the GRNN model has good prediction results during the testing process in terms of both RMSE scores and training times.
• Corresponding author: Ms Wiem Belhedi
  Affiliation: GREYC, Caen Normandie University
  Country: France
  e-mail: