An Analysis of Data Compression Algorithms in the Context of Ultrasonic Bat Bioacoustics

Abstract: Audio data compression seeks to reduce the size of sound files, making them easier to store and transfer, and is thus a highly valued tool for those working with large sets of audio data. For example, some biologists work with audio recordings of bats, which are well known for their frequent use of ultrasonic echolocation, and so these biologists can accrue massive amounts of high frequency audio data. However, as many methods of audio compression are designed to specialize in the more common range of frequencies, they are not able to sufficiently compress bat audio, and many bat biologists instead work without compressing their data at all. This paper investigates the desiderata of a data compression method in the context of bat biology, experimentally compares several modern data compression algorithms, and discusses their pros and cons in terms of their potential use across various relevant contexts. The paper concludes by suggesting the algorithm Monkey’s Audio for machines able to handle the higher resource demands it has. Otherwise, FLAC and WavPack yield similar size reduction rates at a significantly faster speed while being less resource intensive. Of note is the interesting result produced by the algorithm 7-ZIP PPMd Solid, which achieved consistently outstanding results within a single dataset, but its generalizability has yet to be determined.