MARLIN: A Near Real-Time Adaptive Framework for Intelligent Marine Acoustic Monitoring

Passive acoustic recorders are valuable tools for monitoring the marine environment, yet bandwidth limitations have historically prevented real-time streaming of acoustic data to shore for in situ analysis of the soundscape. Advancing the role passive acoustic monitoring (PAM) plays within decision-making pipelines requires adaptive hardware solutions that can deliver ecologically and environmentally critical information in a timely manner. By integrating intelligent detection and classification models directly onto hardware, such platforms could significantly expand the operational capabilities of marine acoustic monitoring. Here the authors present MARLIN (maritime, acoustic, real-time, learning, information, and notification), an adaptive hardware–software framework which enables near real-time streaming of broadband acoustic data, and on-device cutting-edge intelligence for detecting and classifying marine sound sources. The modular architecture of MARLIN supports a range of low-power, single-board computers (edge processing devices) permitting the execution of machine learning algorithms directly on the platform for continuous access to the marine soundscape. Configured to use multiple VPN-encrypted communication pathways, radio links, and Iridium satellites allowing for discrete, accurate real-time operation. A platform independent architecture for coastal monitoring tasks is presented, with a flexible software interface so that users may make use of MARLIN AI, or plug-in custom-built algorithms for personal use cases. In this chapter, the authors describe the hardware and software functionality of MARLIN and demonstrate its capabilities for acoustic monitoring in diverse habitats.