{"id":{"repo_id":"uiuc","oai_identifier":"oai:www.ideals.illinois.edu:2142/127148"},"canonical_url":"https://search.dev.ndltd.org/etd/uiuc/oai:www.ideals.illinois.edu:2142/127148","repository":{"repo_id":"uiuc","name":"University of Illinois - Urbana-Champaign","base_url":"https://www.ideals.illinois.edu/oai-pmh"},"display":{"title":"Environmental adaptation in data-driven underwater acoustic signal processing","abstract":"Submission original under an indefinite embargo labeled 'Open Access'. The submission was exported from vireo on 2025-03-28 without embargo terms","abstract_html":"Submission original under an indefinite embargo labeled 'Open Access'. The submission was exported from vireo on 2025-03-28 without embargo terms","abstract_has_math":false,"creators":["Kari, Dariush"],"institution":"University of Illinois at Urbana-Champaign","degree_name":"Ph.D.","degree_level":"Dissertation","degree_discipline":"Electrical & Computer Engr","degree_department":null,"school":null,"contributors":["Singer, Andrew C","Smaragdis, Paris","Oelze, Michael L","Schwing, Alexander","Hasegawa-Johnson, Mark"],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2024,"date_issued":"2024-09-05","date_published":"2024-09-05","updated_at":"2026-07-22T22:25:03Z","subjects":["Underwater Acoustics","Underwater Acoustic Localization","Data-driven Localization","Gradient-based Localization","Domain Adaptation","Generalization","Environmental Mismatch"],"languages":["en","eng"],"rights":["Copyright 2024 Dariush Kari"],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"https://hdl.handle.net/2142/127148","outbound_label":"Handle","outbound_source":"dc:identifier"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor","label":"Contributor","values":["Singer, Andrew C","Smaragdis, Paris","Oelze, Michael L","Schwing, Alexander","Hasegawa-Johnson, Mark"]},{"key":"dc:creator","label":"Author","values":["Kari, Dariush"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date","label":"Dc Date","values":["2024-09-05","2024-12"]},{"key":"dc:type","label":"Dc Type","values":["text","Thesis"]},{"key":"thesis:degree_discipline","label":"Discipline","values":["Electrical & Computer Engr"]},{"key":"thesis:degree_level","label":"Degree Level","values":["Dissertation"]},{"key":"thesis:degree_name","label":"Degree Name","values":["Ph.D."]},{"key":"thesis:institution_name","label":"Thesis Institution Name","values":["University of Illinois at Urbana-Champaign"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Underwater Acoustics","Underwater Acoustic Localization","Data-driven Localization","Gradient-based Localization","Domain Adaptation","Generalization","Environmental Mismatch"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language","label":"Dc Language","values":["en","eng"]},{"key":"dc:rights","label":"Dc Rights","values":["Copyright 2024 Dariush Kari"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["https://hdl.handle.net/2142/127148"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description","label":"Description","values":["Submission original under an indefinite embargo labeled 'Open Access'. The submission was exported from vireo on 2025-03-28 without embargo terms","The student, Dariush Kari, accepted the attached license on 2024-09-03 at 11:17.","The student, Dariush Kari, submitted this Dissertation for approval on 2024-09-03 at 11:42.","This Dissertation was approved for publication on 2024-09-05 at 16:04.","DSpace SAF Submission Ingestion Package generated from Vireo submission #21211 on 2025-03-28 at 14:25:00","The last decade has witnessed unprecedented advances in machine learning, especially deep learning algorithms, that can be leveraged for underwater acoustic (UWA) signal processing tasks. However, a major obstacle on the way to exploiting the full potential of data-driven algorithms in UWA signal processing is the limited real data available for training deep learning models, which is due to the high costs of data collection in such environments, as well as the ocean variability which hinders generalization of models trained for one environment to other environments. While the generalization problem has been studied from different aspects and for various problems, it has yet to be deeply explored in UWA applications. This thesis investigates the generalization in UWA problems with localization as a downstream task. To this end, we consider both forward modeling, where we seek models that can replicate the acoustic field in an area of interest, and inverse modeling, where we seek solutions to infer the location of an acoustic source from the recorded acoustic signal. To improve the generalization performance, we use adaptation approaches, whose results, while being general enough and able to be applied to other domains, leverage the physics of UWA propagation to reduce the need for training data. In the forward modeling part, we introduce a new deep learning architecture for ray-based data-driven UWA wave propagation, which can be adapted to different environments. While this model can be used for a general UWA propagation scenario, we show that in simpler models, such as shallow water and short ranges or in confined spaces like pools, this model can be considerably simplified and efficiently used for different localization scenarios, such as single sensor setups or simultaneous source and wall localization. Our proposed approach exploits the tools that have been developed for deep learning optimizations and achieves the Cramer-Rao lower bound in high-SNR localization tasks. In addition, we provide an adaptation method to combat the effects of environmental mismatches and the conditions under which the adaptation works satisfactorily. In the inverse modeling part, we first explore deep learning (DL) models uncertainty quantification in UWA localization problems and observe the increase in uncertainty when there is an environmental mismatch. Then, we provide a solution for environmental adaptation by developing a new localization process parallel to the DL model and combining the predictions from this process and the DL model to resolve the ambiguities (uncertainties) of the DL model. Empirical evaluations over the synthetic data and SwellEx-96 data prove the efficacy of the proposed adaptation method. In addition, inspired by the invariant feature extraction, we study several effective ways of using data from several different environments to train inverse models in order to enhance the generalization performance."]},{"key":"dc:format","label":"Dc Format","values":["application/pdf"]},{"key":"dc:title","label":"Title","values":["Environmental adaptation in data-driven underwater acoustic signal processing"]}]}],"canonical_facts":{"dc:contributor":["Singer, Andrew C","Smaragdis, Paris","Oelze, Michael L","Schwing, Alexander","Hasegawa-Johnson, Mark"],"dc:creator":["Kari, Dariush"],"dc:date":["2024-09-05","2024-12"],"dc:description":["Submission original under an indefinite embargo labeled 'Open Access'. The submission was exported from vireo on 2025-03-28 without embargo terms","The student, Dariush Kari, accepted the attached license on 2024-09-03 at 11:17.","The student, Dariush Kari, submitted this Dissertation for approval on 2024-09-03 at 11:42.","This Dissertation was approved for publication on 2024-09-05 at 16:04.","DSpace SAF Submission Ingestion Package generated from Vireo submission #21211 on 2025-03-28 at 14:25:00","The last decade has witnessed unprecedented advances in machine learning, especially deep learning algorithms, that can be leveraged for underwater acoustic (UWA) signal processing tasks. However, a major obstacle on the way to exploiting the full potential of data-driven algorithms in UWA signal processing is the limited real data available for training deep learning models, which is due to the high costs of data collection in such environments, as well as the ocean variability which hinders generalization of models trained for one environment to other environments. While the generalization problem has been studied from different aspects and for various problems, it has yet to be deeply explored in UWA applications. This thesis investigates the generalization in UWA problems with localization as a downstream task. To this end, we consider both forward modeling, where we seek models that can replicate the acoustic field in an area of interest, and inverse modeling, where we seek solutions to infer the location of an acoustic source from the recorded acoustic signal. To improve the generalization performance, we use adaptation approaches, whose results, while being general enough and able to be applied to other domains, leverage the physics of UWA propagation to reduce the need for training data. In the forward modeling part, we introduce a new deep learning architecture for ray-based data-driven UWA wave propagation, which can be adapted to different environments. While this model can be used for a general UWA propagation scenario, we show that in simpler models, such as shallow water and short ranges or in confined spaces like pools, this model can be considerably simplified and efficiently used for different localization scenarios, such as single sensor setups or simultaneous source and wall localization. Our proposed approach exploits the tools that have been developed for deep learning optimizations and achieves the Cramer-Rao lower bound in high-SNR localization tasks. In addition, we provide an adaptation method to combat the effects of environmental mismatches and the conditions under which the adaptation works satisfactorily. In the inverse modeling part, we first explore deep learning (DL) models uncertainty quantification in UWA localization problems and observe the increase in uncertainty when there is an environmental mismatch. Then, we provide a solution for environmental adaptation by developing a new localization process parallel to the DL model and combining the predictions from this process and the DL model to resolve the ambiguities (uncertainties) of the DL model. Empirical evaluations over the synthetic data and SwellEx-96 data prove the efficacy of the proposed adaptation method. In addition, inspired by the invariant feature extraction, we study several effective ways of using data from several different environments to train inverse models in order to enhance the generalization performance."],"dc:format":["application/pdf"],"dc:identifier":["https://hdl.handle.net/2142/127148"],"dc:language":["en","eng"],"dc:rights":["Copyright 2024 Dariush Kari"],"dc:subject":["Underwater Acoustics","Underwater Acoustic Localization","Data-driven Localization","Gradient-based Localization","Domain Adaptation","Generalization","Environmental Mismatch"],"dc:title":["Environmental adaptation in data-driven underwater acoustic signal processing"],"dc:type":["text","Thesis"],"thesis:degree_discipline":["Electrical & Computer Engr"],"thesis:degree_level":["Dissertation"],"thesis:degree_name":["Ph.D."],"thesis:institution_name":["University of Illinois at Urbana-Champaign"]},"updated_at":"2026-07-22T22:25:03Z"}