SongBeam: an automated recorder using beamforming to make high-quality recordings

Recording bird song can provide information about birds’ presence, behaviour and ecology. As such, automated recording units are excellent data collection tools for bioacousticians and ecologists. How- ever, in contrast with directional recorders, the quality of the omnidirectional microphones on ARU’s often puts limits on the usability of the recordings.Here we present a novel song recorder that contains an array of four microphones, the recordings of which can be used for acoustic beamforming to “focus” in on the loudest bird at any one moment. This produces higher-quality recordings with decreased environmental noise and reverberation. The recorder can be programmed to record automatically, and due to its low battery consumption can be deployed in the field for up to months, depending on the recording schedule. We tested the performance of these recorders for 2 different applications: 1) qualitative bird song research and 2) species identification using a species recognition algorithm.We found that beamforming can provide a directional output from a passive recorder. It removed a large proportion of noise and reverberation from the recordings, which benefitted the fine-scale analysis of these songs. Moreover using the beamformed recordings improved the detection rate of a species detection algorithm.SongBeam enables researchers to make high-quality recordings, at a low cost and a low consumption of time and resources. With a low cost of US$100 per unit, SongBeam can be used in large-scale and long-term surveys. These high-quality recordings are not only beneficial for the study of bird song itself but also greatly improve species detectability for biomonitoring and species diversity studies.