Unsupervised acoustic classification of individual gibbon females and the implications for passive acoustic monitoring

Abstract

Passive acoustic monitoring (PAM) has the potential to greatly improve our ability to monitor cryptic yet vocal animals. Advances in automated signal detection have increased the scope of PAM, but distinguishing between individuals—which is necessary for density estimation—remains a major challenge. When individual identity is known, supervised classification techniques can be used to distinguish between individuals. Supervised methods require labelled training data, whereas unsupervised techniques do not. If the acoustic signals of individuals are sufficiently different, the number of clusters might represent the number of individuals sampled. The majority of applications of unsupervised techniques in animal vocalizations have focused on quantifying species‐specific call repertoires. However, with increased interest in PAM applications, unsupervised methods that can distinguish between individuals are needed. Here we use an existing dataset of Bornean gibbon female calls with known identity from five sites on Malaysian Borneo to test the ability of three different unsupervised clustering algorithms (affinity propagation, K‐medoids and Gaussian mixture model‐based clustering) to distinguish between individuals. Calls from different gibbon females are readily distinguishable using supervised techniques. For internal validation of unsupervised cluster solutions, we calculated silhouette coefficients. For external validation, we compared clustering results with female identity labels using a standard metric: normalized mutual information. We also calculated classification accuracy by assigning unsupervised cluster solutions to females based on which cluster had the highest number of calls from a particular female. We found that affinity propagation clustering consistently outperformed the other algorithms for all metrics used. In particular, classification accuracy of affinity propagation clustering was more consistent as the number of females increased, and when we randomly sampled females across sites. We conclude that unsupervised techniques may be useful for providing additional information regarding individual identity for PAM applications. We stress that although we use gibbons as a case study, these methods will be applicable for any individually distinct vocal animal.

Passive Acoustic Monitoring and Automatic Detection of Diel Patterns and Acoustic Structure of Howler Monkey Roars

“…Our study provides additional evidence regarding the utility of PAM and automatic acoustic detection as useful tools to study patterns of activity of primates across their diel activity cycle [22,23,45,46]. PAM coupled with automatic detection of calls allowed us to monitor larger areas for longer periods and gain new insight into the behavioral ecology of this species.…”

“…Emerging new technologies and techniques, such as passive acoustic monitoring (PAM) and automatic acoustic detection of calls [20][21][22], greatly facilitate studying soundproducing animals at night and have the potential to advance both behavioral ecology and conservation practices [23][24][25]. Recently, researchers have used PAM and automatic classifiers to decipher diel and annual vocal cycles of black-and-gold howlers and mantled howlers [5,26] and expanded upon the number of acoustic features analyzed for black howlers (Alouatta pigra) and mantled howlers [6,7].…”

Identifying unknown Indian wolves by their distinctive howls: its potential as a non-invasive survey method

“…This study revealed that the number of wolves present in the recordings could be determined from their howls and the individuality information is sufficient for supervised population estimation through CMR techniques 7 , 25 , 27 , 30 . Therefore, wolves recorded in one location can be acoustically recaptured at another location, and we can identify them individually.…”

“…There are two ways to identify individual wolves or packs—supervised clustering and unsupervised clustering. While supervised clustering requires a set of known training data and cluster validation is straightforward, unsupervised clustering requires ground-truthing before it can be used to monitor populations at a survey level and does not allow individual level CMR or tracking 30 .…”

“…The conventional CMR model requires identification at the individual level 27 , 28 , but it provides a robust population estimation 28 and much additional information such as home-range, survival rate, animal movement pattern, and population viability analysis 29 . However, the difficulty of successfully identifying unknown individuals from their calls has prevented its use for more species, though new techniques are being developed for some species, including the use of unsupervised classifiers to cluster calls 30 . Here, we explore the potential of identifying individuals through supervised classification from their vocal features to potentially improve identification to the point where CMR surveys are possible for an elusive and wide-ranging species.…”