Representing the complexity of natural and human processes in the maritime environment requires the collection and processing of large heterogeneous data sets. Due to the scarcity of sensing resources, information collection needs to be guided by intelligent, agile processes. Therefore, raw heterogeneous data sets need to be standardized and processed locally at the sensing node to reduce communication and computational load associated with transmitting data at a fusion and decision support center. This work presents an IoT framework for maritime applications that consists of two independent, yet compatible hardware designs. One provides maritime data standardization to enable interoperability of ocean sensing systems, and the other provides information acquisition agility to enable efficient allocation of limited edge node resources. An application for ocean sound classification using signal decomposition, suitable for edge processing on-board of IoT systems, is provided as an example of the use of the framework. Three different edge processing implementations are presented and the corresponding performance results are reported and compared.