Data Cleaning applied to Ecoacoustic Datasets: A Literature Review

Passive acoustic monitoring has become an essential tool for understanding biodiversity, enabling detailed analyses of soundscapes across various ecosystems. However, the large volume of data collected presents a significant challenge, particularly due to the difficulty of identifying and correcting errors and irrelevant data, such as sounds resulting from external interference that are not characteristic of the soundscape. Including such data in the analysis may compromise the accuracy of the results and lead to misleading interpretations. Therefore, it is essential to adopt a data cleaning procedure to detect and correct errors occurring during the acquisition process, thereby improving the quality of the data for subsequent analyses. This paper presents a literature review on data cleaning techniques applied to ecoacoustic datasets, exploring methods for identifying and removing outliers, corrupted records, and non-characteristic noise in the soundscape. The findings of this review will provide a foundation for selecting the most suitable techniques in the development of an Artificial Intelligence system aimed at assessing environmental degradation in the Brazilian semi-arid region. By integrating advanced data cleaning strategies into soundscape modeling, this study contributes to improving the accuracy of analyses and the effectiveness of ecoacoustic monitoring.