Agile Edge Classification of Ocean Sounds

Abstract

The maritime environment is characterized by a scarcity of resources of power, sensing, processing, and communications. The resource constraints impose limitations in information acquisition which involves data collection and data processing to yield meaningful statistics. The contribution of this work is on custom software and hardware methods for low power, low data-rate processing for the application of classification of ocean sounds. The combination of light processing software and custom hardware allow the development of efficient cyber-physical maritime IoT systems. A simulation-based study is provided to evaluate the ability of the software method for agile learning of features for ocean sounds classification. In addition, a practical implementation on a custom hardware emulator is provided to demonstrate the potential of the method to classify ocean sounds on low power, inexpensive seaborne IoT nodes.