Abstract

This paper describes an algorithm to generate a Sound-Time Lapse (STL), i.e., audio capsules or acoustic summaries that aim to reliably represent continuous and long audio signals recorded in wetlands from the Chilean city of Valdivia. The algorithm's input is a long duration audio signal (> 20 hours), from which representative chunks are extracted and merged with minimum audible distortion to produce a brief audio file (4-6 minutes). The algorithm is a flexible and adaptable tool, its input parameters can be adjusted by the user to highlight specific portions of the original long recordings. Modern libraries focusing on high performance computing were considered for the implementation of the algorithm. The project aims to highlight the importance of the natural heritage of Valdivia's wetlands and bring them closer to the general public. Additionally , we recognize the opportunity to use STLs in scientific works so as to improve our understanding on biological diversity present in the natural sound composition of wetlands.