Applying machine learning to primate bioacoustics: Review and perspectives

Applying machine learning to primate bioacoustics: Review and perspectives

Abstract

This paper provides a comprehensive review of the use of computational bioacoustics as well as signal and speech processing techniques in the analysis of primate vocal communication. We explore the potential implications of machine learning and deep learning methods, from the use of simple supervised algorithms to more recent self-supervised models, for processing and analyzing large data sets obtained within the emergence of passive acoustic monitoring approaches. In addition, we discuss the importance of automated primate vocalization analysis in tackling essential questions on animal communication and highlighting the role of comparative linguistics in bioacoustic research. We also examine the challenges associated with data collection and annotation and provide insights into potential solutions. Overall, this review paper runs through a set of common or innovative perspectives and applications of machine learning for primate vocal communication analysis and outlines opportunities for future research in this rapidly developing field.

One INTRODUCTION |

Acoustic communication can be observed in many animal species and offers a diverse set of cues in the study of their behavior as well as a prolific insight for the monitoring of their activity. Primates are certainly no exception in this context, and the study of their vocalizations has been of great interest for the scientific community in recent years. Directly following their success in processing speech and audio, deep learning models were introduced to the field of computational bioacoustics through the ever-growing availability of data sets allowed by technical advances in data recording, sharing and storage. This led to the now widely spread passive acoustic monitoring approach and, in turn, to an increasing need for efficient automated workflows in addition to hand-made annotations and analysis. This can be seen as a slight change of paradigm in the way primate vocal communication is treated and understood by researchers, which is also fairly recent and prone to evolve. In fact, deep learning methods developed for speech, audio or image processing as we currently understand them were seldom mentioned in computational bioacoustics reviews until Stowell. Fifteen years after the exploratory perspective paper from Zimmermann et al., one of the first studies mentioning the direct use of artificial neural networks applied to primate vocalizations was carried out by Pozzi et al. Besides that, simpler machine learning approaches developed for the processing of large unsegmented passive acoustic monitoring data sets were not explored, to our knowledge, before the work by Kalan et al. The computational analysis of primate vocal communication systems is thus a young and rapidly growing field of study.

We hereby present a concise survey of the latest trends and approaches in machine learning applied to primate vocal communication research. In this perspective, we carefully selected experiments, mostly published in the last three years, with additional earlier papers that we deemed interesting for a contextualized discussion. After introducing the key concepts and the general approach found in the literature, we describe three different approaches and the type of results they can provide to better understand primate vocal communication or to be used as monitoring tools. We then put forward different perspectives following the development of high-performance weakly supervised acoustic models and their potential use in primate communication research. Finally, we discuss data availability and ongoing efforts in collecting and sharing new exploitable data sets.