In recent years, there has been a significant increase in research on animal sounds, thanks to advancements in recording equipment and analysis techniques. These developments have provided valuable insights into various aspects of animal behavior, population distribution, taxonomy, and anatomy.
However, a new study published in Ecology and Evolution highlights the limitations of one of the most commonly used methods for analyzing animal sounds. These limitations have led to disagreements regarding a whale song in the Indian Ocean and animal calls on land.
The study introduces a new method that can overcome these limitations and reveal previously hidden details of animal calls. This breakthrough has the potential to drive future advancements in animal sound research.
Understanding whale behavior, population distribution, and the impact of human-made noise is crucial for conservation efforts, especially considering that more than a quarter of whale species are listed as vulnerable, endangered, or critically endangered. However, studying these creatures, which spend most of their time hidden in the vast open ocean, poses significant challenges. Analyzing whale songs can provide vital clues in this regard.
Traditionally, analyzing animal sounds involved generating visualizations called spectrograms, which provide a better understanding of the sound’s characteristics. Spectrograms show the temporal details (when the energy in the sound occurs) and spectral details (at what frequency) of a sound. Carefully inspecting these spectrograms and measuring them with other algorithms allows for a deeper analysis and serves as a key tool for communicating research findings.
However, spectrograms generated using the most common method known as the Short-Time Fourier Transform (STFT) have limitations. They cannot accurately visualize all the temporal and spectral details of a sound simultaneously. This means that STFT spectrograms sacrifice either temporal or spectral information.
This limitation becomes more pronounced at lower frequencies, making it problematic when analyzing sounds produced by animals like the pygmy blue whale, whose song is so low that it approaches the lower limit of human hearing.
The study’s lead author, along with co-author Tracey Rogers, compared the STFT method to newer visualization methods. They tested these methods using synthetic test signals and recordings of various animals, including pygmy blue whales, Asian elephants, cassowaries, and American crocodiles.
One of the methods they tested was a new algorithm called the Superlet transform, which they adapted from its original use in brain wave analysis. The researchers found that the Superlet transform produced visualizations with up to 28% fewer errors compared to other methods.
The Superlet transform proved particularly effective when applied to animal sounds. For example, it resolved a debate surrounding the Chagos pygmy blue whale song, accurately showing it as “pulsed” rather than “tonal.” Similarly, when visualizing Asian elephant rumbles, the Superlet transform revealed pulsing that had been mentioned in the original description but was absent from later descriptions and spectrograms.
The Superlet transform also uncovered previously unreported temporal details in the calls of the southern cassowary and the roars of American crocodiles, which were not visible in previous spectrograms.
While these findings are preliminary and based on a single recording, they open up new avenues for future research. To make the Superlet transform more accessible to researchers in ecology, biology, and veterinary science, the team developed a free and user-friendly open-source software app.
Overall, this study highlights the limitations of current methods for analyzing animal sounds and introduces a promising alternative. The Superlet transform has the potential to revolutionize animal sound research by providing more accurate and detailed visualizations, leading to a better understanding of animal behavior and contributing to conservation efforts.