Project: Automated bird song detection
|
Description: The project aims to develop algorithms and tools for automatic analysis of avian acoustics. In tropical forests, because of dense vegetation, birdwatchers typically identify species by means of their calls. Avian diversity in a particular region can be estimated in this manner. In recent times, the use of automated or semi-automated recording devices to capture bird calls is becoming popular. The large volume of data captured by these devices is impractical for human listeners to analyse. Thus, automatic algorithms to perform tasks like species identification and species detection will be a valuable resource to ecologists.
The aim of the project is two fold: to develop low-cost and field-deployable recording hardware, and to develop automatic algorithms for their analysis. The recording devices will be based around a microcontroller, have redundant power supplies, and will store data on a local device like a flash memory card. They will also have weatherproof casing to withstand high variations in the environment. Concurrently, algorithms for birdcall processing will be developed. The development of these algorithms involve many challenges, such as the complex acoustic environment where the recordings are made. This includes, for example, overlapping vocalisations made by several birds, the intra-species call variability, and background sounds made by other animal or non-animal sources. The hardware, algorithms and tools developed by the project will be useful for ecologists in the country to monitor the long-term avian biodiversity of a given region. |
Funding: SERB
Team:
|
