Link to Research Paper : https://docs.google.com/document/d/1Ayii5WDcB4mUkQgFyC-pIf5Md8ry5lCJtiFruJcaKkw/edit?usp=sharing
Abstract
Earth's biodiversity is increasingly under threat as sources of anthropogenic change (resource extraction, land use change, and climate change) continue to drastically shift our landscapes. One of the species most affected by human interference is the African forest elephants (Loxodonta Cyclontis); in the last decade itself, the population has fallen more than 60%, leaving African elephants in an endangered status. In this state, population monitoring is critical to ensure the elephants survival, however, manually monitoring tropical forests is intractable due to the lack of resources and effectiveness. Possible methods of achieving the monitoring of the elephant populations include placing GPS collars around elephant necks and monitoring their presence using drones, however, both of these methods are impractical due to the lack of resources. In light of these problems, it is clear that the development of a feasible and reliable method of monitoring is necessary. The application of machine learning in bioacoustics allows us to acoustically determine the presence of elephants in a certain area by hacking into their private infrasound communication channels. Using state-of-art techniques in artificial intelligence and a large, professionally labeled dataset of passive acoustic recordings of the African Forest Elephant, I have developed a model that can successfully detect an elephant's presence with a 99.3% accuracy. This model was then deployed into an IoT device that can communicate with wild-life conservatives in real-time. This advancement could potentially revolutionize wildlife conservation and save countless animal lives from endangerment.