Abstract
Unmanned aerial vehicles are evolving rapidly and being deployed in increasingly varied contexts. As their roles broaden, acoustics is gaining prominence as a complementary, passive source of information, especially where other sensing modalities face constraints. To situate the field and guide consistent evaluation and deployment, there is a need to consolidate methods, datasets, and open challenges. Motivated by these trends, this review provides a comprehensive overview of algorithms and applications in unmanned aerial vehicle acoustics, highlighting the increasing importance of sound-based approaches in advancing this field. It explores the use of signal processing, machine learning, and deep learning to analyze and process acoustic data. Additionally, the review examines publicly available acoustic recording databases specific to unmanned aerial vehicles, describing their structure, scope, and limitations. The study emphasizes the importance of robust, diverse, and scalable databases in bridging the critical link between algorithm development and dataset availability. Furthermore, it identifies and discusses future research directions, focusing on strategies to integrate sound and unmanned aerial vehicle technologies. Together, these elements clarify the methodological landscape and highlight the urgent need for well-annotated, diverse, and scalable datasets to support transparent benchmarking and deployment.
Funding Information
  • Project of the Military University of Technology titled: New Neural Network Architectures for Signal and Data Processing in Radiocommunications and Multimedia (UGB 22-054/2025)

This publication has 164 references indexed in Scilit: