Answers are grounded in this paper and cite their sources.
In inland waterway maritime, AIS and VHF voice communication, due to overlapping speakers, strong non-stationary noise hinders automation and intelligent supervision. This study proposes a method for predicting the berthing behaviour intentions of ships by integrating AIS data with maritime VHF voice feature selection and recognition. The method employs threshold-based screening and preprocessing to remove invalid audio components, followed by denoising and enhancement using a denoising convolutional neural network. An integrated attention mechanism enables the model to focus on key information related to vessel intentions. Experimental results show that the proposed approach reduces the character error rate for keyword recognition from 70% to 41.27% after AIS big data integration denoising, and to 20.54% with domain-specific training—substantially outperforming traditional methods. This research significantly enhances the automation and accuracy of vessel traffic service operations, offering robust technical support for the safety of maritime communications management.
No takes yet. Share an insight, caveat, or question.
Bai et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: