EcoEar Detection of Koala Calls

The Challenge

Koalas are a threatened species in NSW. As there are numerous programs set out to revive the population, there is a need to better understand the exact number of koalas in NSW over time. Current methods include manual field surveys which have limited coverage and provide logistical challenges. The NSW SBIR program called for new technologies to increase this range and accuracy of detection.  

Biodiversity Monitoring Services have acknowledged that acoustic surveys offer large coverage of koala calls in a format that allows operators to ‘set and forget’ the sensors. The challenge comes from the generation of enormous data sets - most of which are empty recordings. These then require manual analysis from an ecologist, increasing cost and time, rendering them impractical and resulting in vast datasets that are difficult to curate and use for decision-making.

The Solution

Biodiversity Monitoring Services proposes a solution that will address both power efficiency, data storage and data analysis challenges. Researchers at the International Centre for Neuromorphic Systems University of Western Sydney (in conjunction with Biodiversity Monitoring Services and Neurabuild) have developed the EcoEar, an AI-powered edge device that is only triggered by koala calls and otherwise stays dormant. This results in an extremely power-efficient device that can stay in the environment for months on end and only detect and record data when Koala calls are made. This is in contrast to previous devices that generate terabytes of data representing weeks or months of raw unlabelled audio, the EcoEar returns a labelled list of detected call events that ecologists can immediately use. The EcoEar streamlines ecological monitoring by directly providing ecologists with targeted, actionable data making large-scale monitoring feasible and efficient.

Such a system has been demonstrated to detect dingo calls in a previous study at the University of Western Sydney, and this project repurposed this technology to koala populations. To do this, a new artificial intelligence model was trained using koala call data and implemented in the hardware of this novel sensing device. 

Industry Partners

University Partners