Evolution of performance metrics in the sonar world

Abstract

The performance of sonar technology, including passive acoustic monitoring, environmental impact assessment and the monitoring of man-made noise sources, is typically characterized by the through-water detection range. This single performance measure depends on the characteristics of the sound sources, the underwater environment and sensing capabilities. Measuring sonar performance is considered critical in the deployment, operation and development of sonar systems, otherwise one is danger of investing in the wrong technology and/or operating this technology incorrectly. It is however already known that using detection range by itself may be insufficient, for instance where the detection and classification tasks are inseparable and counter-detection needs to be taken into account. This talk will describe the different aspects of sonar performance that need to be considered and correctly assessed. In particular, ongoing work to identify better metrics for detection, classification and localization will be outlined. This work is looking initially to borrow metrics from other domains and applications, both for speed of development and retention of cross-domain compatibility, in particular from the area of Machine Learning. Finally, this talk will identify the major areas where further work may still be required.