dc.description.abstract |
Kenya's rich biodiversity faces a number of threats including human
encroachment, poaching and climate change. Since Kenya is a developing country,
there is need to manage the sometimes competing interests of development, such as
infrastructure development, and conservation. To achieve this, tools to effectively
monitor the state of Kenya's various ecosystems are essential. In this paper we
propose a biodiversity monitoring software tool that integrates acoustic indices of
biodiversity, recognition of species of interest based on their vocalizations and
acoustic census. This tool can be used by non-experts to determine the current state
of their ecosystems by monitoring the state of bird species that serve as indicator
taxa and whose abundance is related to the abundance of other terrestrial vertebrates
including the “big five”. The tool we propose exploits state-of-the art advances in
signal processing and machine learning to perform biodiversity monitoring, bird
species detection and census in a joint framework. Using publicly available data we
demonstrate how current acoustic indices of biodiversity can be improved by
incorporating machine learning based audio segmentation algorithms. We also show
how open source toolkits can be used to build bird species recognition systems. Code
to reproduce the experiments in this paper is available on Github at
https://github.com/ciiram/BirdPy. |
en_US |