BROWSE
Detail
Conference Paper
Architecture of an AI-Based Distress Detection and Situation Assessment System for Maritime Emergency Response
Cited - time in 
Share 
Share
- Authors
- Jin-Hyuk Song, Minjung Lee, Eunjung Kwon
- Issue Date
- 2026-08
- Citation
- International Conference on Mobile ∙ Military ∙ Maritime IT Convergence (ICMIC) 2026, pp.1-4
- Publisher
- IEEE
- Language
- English
- Type
- Conference Paper
- Abstract
- Maritime emergency reception is a mission-critical process in which the rapid and accurate identification of distress situations directly affects rescue timing and survival outcomes. Current operational workflows still rely heavily on human listening and manual judgment across very high frequency (VHF), single sideband (SSB), digital selective calling (DSC), satellite signals, and telephone reports. As a result, emergency operators remain vulnerable to misrecognition caused by background noise, radio interference, overlapping speakers, and fatigue resulting from long-duration monitoring. In addition, multiple digital distress signals, such as DSC, emergency position-indicating radio beacon (EPIRB), V-Pass, and e-Navigation events, are generated in the maritime environment, but a single signal alone is often insufficient to determine whether an actual accident has occurred. This limitation leads to unnecessary dispatches, increased operator workload, and inefficient use of rescue resources. To address these issues, this paper proposes an AI-based maritime emergency reception and response system that integrates speech-derived emergency reports and digital distress signals into a unified decision-support framework. Among the many component technologies required for the complete system, this paper presents two core technologies. The first is a maritime-domain named entity recognition (NER) model designed to extract key distress information, such as accident type, location, number of persons, and risk indicators, from low-quality speech-to-text (STT) output. The second is an AI model that estimates the likelihood of a real distress event by analyzing temporal, spatial, and behavioral consistency across heterogeneous digital distress signals. This paper is intended as a system design proposal rather than an evaluation study, and its main contribution lies in presenting an implementable architecture and a technical direction for future deployment and validation in real maritime emergency operations.
- Keyword
- maritime distress, emergency reception, natural language processing, named entity recognition, digital distress signal, multi-modal fusion, decision support
- KSP Keywords
- Assessment system, Background noise, Component technology, Decision-support framework, E-Navigation, Emergency operations, Emergency response, Evaluation study, High frequency(HF), Maritime emergency, Maritime environment