QinetiQ: Passive Sonar Research and Development
Anomify collaborated with QinetiQ on a research and development project to advance passive sonar capabilities.
The core challenge was to enhance the ability to listen for and identify underwater signals without emitting any prior signal - a critical requirement for discreet undersea monitoring. This capability was aimed at identifying marine craft by their unique engine sound signatures and - looking further - even recognizing certain sea creatures. Successfully addressing this would enable comprehensive monitoring of undersea activity and provide crucial information for modifying human maritime behaviour to protect endangered or sensitive marine species when detected.
The collaborative process involved in-depth research and development focused on new techniques for interpreting these complex underwater acoustics. The project recognized that while acoustic sensing (sonar) has been a mainstay, new applications and technological developments are continually emerging. The goal was to move beyond simple detection to sophisticated classification of underwater sound sources within this challenging acoustic landscape. The underwater environment itself presents significant hurdles, being vast and noisy with biological sounds (like shrimp and whales), wind, waves, and existing man-made sounds, and having variable sound propagation based on factors like salinity and temperature.
Anomify’s contribution was the development of a novel system that combined our deep expertise in anomaly detection with our team’s additional specialized understanding of digital audio principles. Our approach allowed for new ways to analyze the subtle and often faint acoustic signatures present in passive sonar data.
We used a combination of spectral and signal-to-noise processing to identify potential candidates for analysis, and then attempted to group signals which had similar patterns of frequency and amplitude change over time, as sounds which could potentially be from the same source. For example - a boat engine will produce a number of frequencies, which when heard statically could appear unrelated, but when observed all rising or falling in tone at the same time, or changing in amplitude in similar ways, we could form a hypothesis that these sounds were coming from the same source. From this we could attempt to group all of the frequencies emanating from a marine craft, and make assumptions about its direction and speed.
The system and its potential were presented by Anomify in a talk to QinetiQ and members of the UK Admiralty and the Ministry of Defence.
As a bonus, we also found that our algorithm was able to track the sounds generated by bees!
Related links: