Automatic Speech Recognition for Search and Rescue: A Dataset Generation Framework for Irish Coast Guard Operations

In Ireland, Maritime Rescue Coordination Centres (MRCCs) receive distress calls and alerts through multiple communication channels, including VHF radio. Irish MRCC staff perform a dual role: operating as Coast Radio Stations and coordinating Search and Rescue (SAR) operations. Rapid and accurate communication with vessels and aircraft is essential in life-threatening maritime incidents. Automatic Speech Recognition (ASR) systems offer the potential to support MRCC watch officers during high-intensity communication periods by automatically transcribing VHF radio calls. This would enable real-time completion of radio logs, allowing officers to focus on maintaining situational awareness and effectively managing SAR coordination tasks. This study evaluates current ASR models, datasets, and fine-tuning methodologies for deployment in this unique operational context. We propose the development of regionally adapted datasets for fine-tuning foundational ASR models. This contribution addresses the specific linguistic and acoustic challenges faced locally by MRCCs with practical ASR deployments in Maritime VHF audio settings.