In this work, we explore machine learning through a model-agnostic feature representation known as braiding, that employs braid manifolds to interpret multipath ray bundles. We generate training and testing data using the well-known BELLHOP model to simulate shallow water acoustic channels across a wide range of multipath scattering activity. We examine three different machine learning techniques—k-nearest neighbors, random forest tree ensemble, and a fully connected neural network—as well as two machine learning applications. The first application applies known physical parameters and braid information to determine the number of reflections the acoustic signal may undergo through the environment. The second application applies braid path information to determine if a braid is an important representation of the channel (i.e., evolving across bands of higher amplitude activity in the channel). Testing accuracy of the best trained machine learning algorithm in the first application was 86.70% and the testing accuracy of the second application was 99.94%. This work can be potentially beneficial in examining how the reflectors in the environment changeover time while also determining relevant braids for faster channel estimation.
Explainable artificial intelligence is gaining wider traction within the machine learning community and its application domains at large for its inherent motivation to explain and interpret large-scale machine interpretation of experimental datasets. We will present cognitive sampling as new way to implement explainable AI for acoustical signal processing. In particular, methods based on geometric signal processing and sparse sensing will be harnessed with machine cognition to interpret, classify and predict information autonomously from large-scale acoustic datasets spanning a wide variety of applications. We will compare the difference in performance between traditional supervised and semi-supervised learning architectures such as deep learning, and ensemble approaches, unsupervised learning networks. We will also present preliminary research on implementing cognitive sampling in machine-directed inverse problem-solving techniques such as autoencoders. The end goal is to discover efficient data encodings that enable hitherto unforeseen feature spaces using optimal or close to optimal sampling strategies. Specific applications will include a variety of acoustical environmental sensing applications involving spectral feature generation and interpretation such as sonar signal processing, undersea multipath channel sensing, as well as feature extraction in complex melodic structures in Indian classical music. [Work funded partially by ONR under Grants N00014-19-1-2436, N00014-19-1-2609, N00174-20-1-0016 and N00014-20-1-2626.]
Robust underwater acoustic channel estimation is critical towards improving communications efforts and enhancing awareness of changing environments. To explore these channels in-depth, machine learning algorithms are developed through feature geometric representations, referred to as “braiding,” to interpret multipath ray bundles within shallow water acoustic channels in two ways. The first application of this work predicts the number of reflections an acoustic signal may undergo through the environment by applying known physical parameters and braid features. The second application explores the importance of a braid feature within the acoustic channel for estimation purposes by using braid path information. Three unique machine learning techniques are trained to predict the applications using a diverse set of shallow water acoustic channels generated through the BELLHOP model. Machine learning models developed for the applications demonstrate high testing accuracies with an accuracy of 86.70% in the first application and an accuracy of 99.94% in the second application. As a further demonstration, braid feature representations and model predictions are used for channel estimation and determining the number of reflections using SPACE08 field data.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.