Visual analysis of bioacoustics annotations in long duration audio data using parallel coordinates

This paper explores the utility of parallel coordinates (PC) to reveal patterns and insights from large annotated bioacoustics events in the Ecosounds audio repository. Working with long environmental acoustic recordings has the issue of data opacity and heterogeneity with increasing volume and therefore requires human-in-the-loop analysis. This process has generated huge amounts of annotated data of various vocalizing species. However, to date there are limited tools and techniques that support visual exploration and analysis of the annotated datasets. The goal of this paper is to explore large amounts of bioacoustics annotations and related acoustic features by using Parallel Coordinates. We demonstrate two distinct usecases. First, we present visual analysis of large bird species call annotations with weather data. Second, we demonstrate how parallel Coordinates can be used for visual analysis and characterization of acoustic features for multiple cohabiting frog species at a given site. Our preliminary exploration shows that Parallel Coordinates can support analysis and exploration of bioacoustics events in long duration environmental recordings. N Furthermore, parallel coordinates can be used for visual characterization of acoustic features or indices that can feed into automated analysis.