@article{M0627F13B,
title = "Acoustic Novelty Detection Using Monte Carlo Dropout and Gaussian Mixture Model in Underwater Acoustic Environments",
journal = "The Journal of Korean Institute of Communications and Information Sciences",
year = "2024",
issn = "1226-4717",
doi = "10.7840/kics.2024.49.12.1702",
author = "Nayeon Kim, Chanjun Chun, Hong Kook Kim",
keywords = "Acoustic Novelty Detection, Monte Carlo Dropout, Gaussian mixture model",
abstract = "In this paper, we propose a method for detecting novel acoustic signals that deviate from the learned data distribution in underwater acoustic environments.
Specifically, we utilize Monte Carlo dropout (MCDO) to quantify the uncertainty in the model's predictions and model the distribution of predictions for normal signals using a Gaussian Mixture Model (GMM). We measure the distance between the estimated GMM and the Gaussian mixture model obtained from the input signal. If this distance exceeds a predefined threshold, the signal is detected as an unlearned novelty signal, indicating that it deviates from the training data distribution. This framework offers a new approach to identifying acoustic anomalies by leveraging model uncertainty and the probabilistic modeling capabilities of GMMs."
}