Standard ecological survey methods are costly to complete. Passive acoustic monitoring of environments offer an advantage in terms of long-term or multiple site analysis. The challenge is in developing robust methods to process the acoustic signals to yield reliable indicators of vocal activity. This project will utilise Machine Learning techniques where the availability of labelled data is often a limiting factor. Therefore, both supervised and unsupervised methods will be considered. Additionally, crowd-sourced collection of data labels may also provide the means to capture sufficient training data, though it’s reliability must be tested. There will also be an opportunity to work with biologists to develop practical survey methodologies to utilise these new algorithms.
Candidates must be from the EU and will need a 1st class or high 2:1 honours degree in a relevant subject such computing, mathematics, engineering or a physical science. As most of the project will require application of Machine learning methods a good understanding of these methods is essential. Additionally, a good understanding of Engineering Mathematics, Digital Signal Processing and Statistics is desirable.