Machine Learning Algorithms for Automatic Classification of Marmoset Vocalizations

Turesson, Hjalmar K.; Ribeiro, Sidarta; Pereira, Danillo Roberto; Papa, João Paulo; Albuquerque, Victor Hugo C. de

doi:10.1371/journal.pone.0163041

Cited by 38 publications

(42 citation statements)

References 53 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Following the collection of additional recordings, one could then test different machine learning algorithms for the automatic classification of vocalization (e.g. artificial neural network [ANN], knearest neighbors [k-NN], random forest classifier; Stowell & Plumbley 2014, Turesson et al 2016 in order to find the most reliable classification method for the long-term acoustic monitoring of the Mediterranean monk seal. The ongoing development of the automatic classification of animal vocalizations has proven its efficacy (Mielke & Zuberbühler 2013, Stowell & Plumbley 2014 and has opened new perspectives/opportunities in developing powerful tools to monitor animal biodiversity.…”

Section: Individual Vocal Signatures and Monitoringmentioning

confidence: 99%

Individual signatures in the vocal repertoire of the endangered Mediterranean monk seal: new perspectives for population monitoring

Charrier

Marchesseau

Dendrinos

et al. 2017

Endang. Species. Res.

View full text Add to dashboard Cite

The Mediterranean monk seal Monachus monachus is considered Endangered by the IUCN, and is the most endangered pinniped in the world. Increasing our knowledge of this species is crucial in order to further our understanding of its social behaviour, but also to develop new methods to monitor and protect it. In many species, acoustic communication plays a major role in social interactions, and vocal signals convey important information about the emitter; understanding the diverse information encoded in vocalizations is helpful in wildlife monitoring. In the present study, we used passive, audio-video surveys to describe the aerial vocal repertoire of the Mediterranean monk seal during the pupping season. An exhaustive analysis was performed on the different call types, and individual vocal signatures were investigated. A total of 5 call types were identified: bark, chirp, grunt, short scream and scream, with bark and scream being the 2 main call types. A discriminant function analysis based on 10 acoustic variables revealed that all call types except grunts can be correctly classified, with an average rate of 86.7%. Furthermore, the individual vocal signature investigated in barks and screams revealed that both call types are individually specific, showing average correct classification rates of 54.2 and 66.1% respectively. Based on these findings, future research should focus on collecting new recordings from wellidentified seals to develop a new passive acoustic monitoring system based on individual identification. This system will enable the evaluation of annual pup production and thus provide essential information on the conservation status of the Mediterranean monk seal in Greece.

show abstract

Section: Individual Vocal Signatures and Monitoringmentioning

confidence: 99%

Individual signatures in the vocal repertoire of the endangered Mediterranean monk seal: new perspectives for population monitoring

Charrier

Marchesseau

Dendrinos

et al. 2017

Endang. Species. Res.

View full text Add to dashboard Cite

show abstract

“…Unlike previous efforts that classify call types on noise-free, pre-segmented audio (Agamaite et al, 2015;Turesson et al, 2016), our network uses raw, noisy spectrograms. The network is thus robust in detecting and classifying in an environment that is noisy, and doesn't require any preprocessing of the input audio stream.…”

Section: Discussionmentioning

confidence: 99%

“…Attribution (which animal vocalized). Automated classification has been done in various species, including rodents (Kobayasi & Riquimaroux, 2012;Soltis, Alligood, Blowers, & Savage, 2012), frogs (Pettitt, Bourne, & Bee, 2012), birds (Giret et al, 2011), bats (Prat, Taub, & Yovel, 2016) and primates (Fuller, 2014;Hedwig, Hammerschmidt, Mundry, Robbins, & Boesch, 2014), including marmosets (Agamaite, Chang, Osmanski, & Wang, 2015;Turesson, Ribeiro, Pereira, Papa, & De Albuquerque, 2016;Zhang, Huang, Gong, Ling, & Hu, 2018). In previous marmoset studies, it was possible to detect the vocalizations by amplitude thresholding of the band-or high-passed audio signal.…”

Section: Background and Related Workmentioning

confidence: 99%

“…As input for classification Agamaite et al extracted 18 acoustic features, chosen by the investigators, in both the time and frequency domain (see Table I in Agamaite et al, 2016). Turesson et al used Linear Predictive Coding filters for feature extraction (Turesson et al, 2016). Zhang et al, (2018) is the most recent work to automatically classify marmoset calls with higher classification accuracy and low frame error rate.…”

Section: Background and Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Deep Convolutional Network for Animal Sound Classification and Source Attribution using Dual Audio Recordings

Oikarinen

Srinivasan

Meisner

et al. 2018

Preprint

View full text Add to dashboard Cite

We introduce an end-to-end feedforward convolutional neural network that is able to reliably classify the source and type of animal calls in a noisy environment using two streams of audio data after being trained on a dataset of modest size and imperfect labels. The data consists of audio recordings from captive marmoset monkeys housed in pairs, with several other cages nearby. Our network can classify both the call type and which animal made it with a single pass through a single network using raw spectrogram images as input. The network vastly increases data analysis capacity for researchers interested in studying marmoset vocalizations, and allows data collection in the home cage, in group housed animals.

show abstract

“…To the best of our knowledge, only one publication has 58 reported on a preliminary method for automated behavioral tracking of marmosets [6]. 59 In contrast to rodents, primates in general have to orient their gaze precisely which 60 makes gaze direction informative of what they pay attention to [33,49]. Further, a 61 striking feature of the gaze behavior of small-headed primates such as marmosets is 62 their rapid head movements.…”

mentioning

confidence: 99%

Head and gaze tracking of unrestrained marmosets

Turesson

Conceicao

Ribeiro

2016

Preprint

Self Cite

View full text Add to dashboard Cite

New technologies for manipulating and recording the nervous system allow us to 1 perform unprecedented experiments. However, the influence of our experimental 2 manipulations on psychological processes must be inferred from their effects on 3 behavior. Today, quantifying behavior has become the bottleneck for large-scale, high 4 throughput, experiments. The method presented here addresses this issue by using deep 5 learning algorithms for video-based animal tracking. Here we describe a reliable 6 automatic method for tracking head position and orientation from simple video 7 recordings of the common marmoset (Callithrix jacchus). This method for measuring 8 marmoset behavior allows for the indirect estimation of gaze within foveal error, and 9 can easily be adapted to a wide variety of similar tasks in biomedical research. In 10 particular, the method has great potential for the simultaneous tracking of multiple 11 marmosets to quantify social behaviors. 12 Introduction 13Recent technological developments allow us to manipulate [3,35,40] and record [7,9,53] 14 the nervous system with historically unprecedented precision and scale. Yet, the 15 psychological relevance of our sophisticated manipulations and large-scale recordings 16 can only be inferred from their effect on behavior. Today, quantifying behavior has 17 become the bottleneck for high-throughput experiments [5,48]. Common practice is to 18 apply standard tests designed to measure psychological constructs such as anxiety and 19 memory, for example, by using the Elevated Plus Maze [34] or the Morris Water 20 Maze [28]. Such testing requires animals to be individually handled, making data 21 acquisition labor intensive, increasing costs and reducing experimental throughput. 22Alternatively, various simple detectors (e.g. capacitance sensors or photo-beams) can be 23 arranged to automatically acquire data at specific sites (e.g. drinking [12] and 24 feeding [10] stations). This type of automatization allows for high throughput, but fails 25 to capture complex or subtle behaviors such as social interactions and gaze behaviors. A 26 more promising approach is to record high-dimensional data from sensor arrays (e.g. 27video) and extract relevant information using computer vision algorithms. This 28 approach has the potential to provide a better characterization of behavior, capable of 29 automatically capturing complex and subtle behaviors, while simultaneously reducing 30 both cost and labor intensity [39]. Raw video frames are composed of a high number of 31 pixels whose values do not straightforwardly correlate with an animal's behavior. In

show abstract

Machine Learning Algorithms for Automatic Classification of Marmoset Vocalizations

Cited by 38 publications

References 53 publications

Individual signatures in the vocal repertoire of the endangered Mediterranean monk seal: new perspectives for population monitoring

Individual signatures in the vocal repertoire of the endangered Mediterranean monk seal: new perspectives for population monitoring

Deep Convolutional Network for Animal Sound Classification and Source Attribution using Dual Audio Recordings

Head and gaze tracking of unrestrained marmosets

Contact Info

Product

Resources

About