Abstract:
Ecosystem monitoring using bats as bio indicators can be achieved through their echolocation recordings. By analysing the content of bats’ recordings, ecologists can infer aspects such as species diversity, population dynamics among others. This information is crucial in assessing ecosystem health. Collecting recordings passively is straightforward by deploying recorders. Drawing inferences from these recordings calls for automatic screening tools to help ecologists detect, localise and characterise bat calls present. We developed an audio processing pipeline to enhance screening of acoustic recordings for bats’ calls detection and localization. The recordings were collected within Dedan Kimathi University of Technology using AudioMoth recorders. Our pipeline leverages simple methods such as median clipping and serves as an initial screening stage before further analysis using sophisticated methods such as machine learning techniques. The pipeline obtained good results and proved effective in detecting and localising bat calls from audio recordings.