Desenvolvimento de um software para rotulação de eventos acústicos

Abstract

With the increasing climate changes, it is important to have mechanisms to verify and identify how such changes affect the environment. In this context, environmental monitoring becomes essential for the preservation of biodiversity. To this end, a viable alternative is to monitor environmental preservation areas to identify the bird species present, as birds are sensitive to climatic variations. In this way, monitoring allows for the analysis of the impact of these changes on the environment, as well as the control of bird migratory flows. A non-invasive alternative for carrying out this monitoring is the automatic identification of species through vocalizations using Machine Learning algorithms. For this, it is necessary to build labeled databases. Doing so without the proper tools can be costly, depending on the size of the dataset. Therefore, this work proposes the development of a system for the manual labeling of acoustic events in partnership with NAPI Biodiversity. Although there are alternatives available on the market, this system is completely free, open-source, and tailored to meet NAPI’s specific usability needs. The main features include the creation of labels on audios rendered in spectrogram format and the export of CSV files containing information about each label. Additionally, it is possible to control the parameters for rendering the audios in spectrogram format.