ORCA-SPOT: An Automatic Killer Whale Sound Detection Toolkit Using Deep Learning

Many morphs: Parsing gesture signals from the noise

Parsing signals from noise is a general problem for signallers and recipients, and for researchers studying communicative systems. Substantial efforts have been invested in comparing how other species encode information and meaning, and how signalling is structured. However, research depends on identifying and discriminating signals that represent meaningful units of analysis. Early approaches to defining signal repertoires applied top-down approaches, classifying cases into predefined signal types. Recently, more labour-intensive methods have taken a bottom-up approach describing detailed features of each signal and clustering cases based on patterns of similarity in multi-dimensional feature-space that were previously undetectable. Nevertheless, it remains essential to assess whether the resulting repertoires are composed of relevant units from the perspective of the species using them, and redefining repertoires when additional data become available. In this paper we provide a framework that takes data from the largest set of wild chimpanzee (Pan troglodytes) gestures currently available, splitting gesture types at a fine scale based on modifying features of gesture expression using latent class analysis (a model-based cluster detection algorithm for categorical variables), and then determining whether this splitting process reduces uncertainty about the goal or community of the gesture. Our method allows different features of interest to be incorporated into the splitting process, providing substantial future flexibility across, for example, species, populations, and levels of signal granularity. Doing so, we provide a powerful tool allowing researchers interested in gestural communication to establish repertoires of relevant units for subsequent analyses within and between systems of communication.

Deep learning based decoding of single local field potential events

How is information processed in the cerebral cortex? In most cases, recorded brain activity is averaged over many (stimulus) repetitions, which erases the fine-structure of the neural signal. However, the brain is obviously a single-trial processor. Thus, we here demonstrate that an unsupervised machine learning approach can be used to extract meaningful information from electro-physiological recordings on a single-trial basis. We use an auto-encoder network to reduce the dimensions of single local field potential (LFP) events to create interpretable clusters of different neural activity patterns. Strikingly, certain LFP shapes correspond to latency differences in different recording channels. Hence, LFP shapes can be used to determine the direction of information flux in the cerebral cortex. Furthermore, after clustering, we decoded the cluster centroids to reverse-engineer the underlying prototypical LFP event shapes. To evaluate our approach, we applied it to both extra-cellular neural recordings in rodents, and intra-cranial EEG recordings in humans. Finally, we find that single channel LFP event shapes during spontaneous activity sample from the realm of possible stimulus evoked event shapes. A finding which so far has only been demonstrated for multi-channel population coding.

Bat2Web: A Framework for Real-Time Classification of Bat Species Echolocation Signals Using Audio Sensor Data

Bats play a pivotal role in maintaining ecological balance, and studying their behaviors offers vital insights into environmental health and aids in conservation efforts. Determining the presence of various bat species in an environment is essential for many bat studies. Specialized audio sensors can be used to record bat echolocation calls that can then be used to identify bat species. However, the complexity of bat calls presents a significant challenge, necessitating expert analysis and extensive time for accurate interpretation. Recent advances in neural networks can help identify bat species automatically from their echolocation calls. Such neural networks can be integrated into a complete end-to-end system that leverages recent internet of things (IoT) technologies with long-range, low-powered communication protocols to implement automated acoustical monitoring. This paper presents the design and implementation of such a system that uses a tiny neural network for interpreting sensor data derived from bat echolocation signals. A highly compact convolutional neural network (CNN) model was developed that demonstrated excellent performance in bat species identification, achieving an F1-score of 0.9578 and an accuracy rate of 97.5%. The neural network was deployed, and its performance was evaluated on various alternative edge devices, including the NVIDIA Jetson Nano and Google Coral.

Animal-borne soundscape logger as a system for edge classification of sound sources and data transmission for monitoring near-real-time underwater soundscape

The underwater environment is filled with various sounds, with its soundscape composed of biological, geographical, and anthropological sounds. Our work focused on developing a novel method to observe and classify these sounds, enriching our understanding of the underwater ecosystem. We constructed a biologging system allowing near-real-time observation of underwater soundscapes. Utilizing deep-learning-based edge processing, this system classifies the sources of sounds, and upon the tagged animal surfacing, it transmits positional data, results of sound source classification, and sensor readings such as depth and temperature. To test the system, we attached the logger to sea turtles (Chelonia mydas) and collected data through a cellular network. The data provided information on the location-specific sounds detected by the sea turtles, suggesting the possibility to infer the distribution of specific species of organisms over time. The data showed that not only biological sounds but also geographical and anthropological sounds can be classified, highlighting the potential for conducting multi-point and long-term observations to monitor the distribution patterns of various sound sources. This system, which can be considered an autonomous mobile platform for oceanographic observations, including soundscapes, has significant potential to enhance our understanding of acoustic diversity.

Development of a machine learning detector for North Atlantic humpback whale song

The study of humpback whale song using passive acoustic monitoring devices requires bioacousticians to manually review hours of audio recordings to annotate the signals. To vastly reduce the time of manual annotation through automation, a machine learning model was developed. Convolutional neural networks have made major advances in the previous decade, leading to a wide range of applications, including the detection of frequency modulated vocalizations by cetaceans. A large dataset of over 60 000 audio segments of 4 s length is collected from the North Atlantic and used to fine-tune an existing model for humpback whale song detection in the North Pacific (see Allen, Harvey, Harrell, Jansen, Merkens, Wall, Cattiau, and Oleson (2021). Front. Mar. Sci. 8, 607321). Furthermore, different data augmentation techniques (time-shift, noise augmentation, and masking) are used to artificially increase the variability within the training set. Retraining and augmentation yield F-score values of 0.88 on context window basis and 0.89 on hourly basis with false positive rates of 0.05 on context window basis and 0.01 on hourly basis. If necessary, usage and retraining of the existing model is made convenient by a framework (AcoDet, acoustic detector) built during this project. Combining the tools provided by this framework could save researchers hours of manual annotation time and, thus, accelerate their research.

Variation in foraging activity influences area-restricted search behaviour by bottlenose dolphins

Area-restricted search (ARS) behaviour is commonly used to characterize spatio-temporal variation in foraging activity of predators, but evidence of the drivers underlying this behaviour in marine systems is sparse. Advances in underwater sound recording techniques and automated processing of acoustic data now provide opportunities to investigate these questions where species use different vocalizations when encountering prey. Here, we used passive acoustics to investigate drivers of ARS behaviour in a population of dolphins and determined if residency in key foraging areas increased following encounters with prey. Analyses were based on two independent proxies of foraging: echolocation buzzes (widely used as foraging proxies) and bray calls (vocalizations linked to salmon predation attempts). Echolocation buzzes were extracted from echolocation data loggers and bray calls from broadband recordings by a convolutional neural network. We found a strong positive relationship between the duration of encounters and the frequency of both foraging proxies, supporting the theory that bottlenose dolphins engage in ARS behaviour in response to higher prey encounter rates. This study provides empirical evidence for one driver of ARS behaviour and demonstrates the potential for applying passive acoustic monitoring in combination with deep learning-based techniques to investigate the behaviour of vocal animals.

ANIMAL-SPOT enables animal-independent signal detection and classification using deep learning

Bioacoustic research spans a wide range of biological questions and applications, relying on identification of target species or smaller acoustic units, such as distinct call types. However, manually identifying the signal of interest is time-intensive, error-prone, and becomes unfeasible with large data volumes. Therefore, machine-driven algorithms are increasingly applied to various bioacoustic signal identification challenges. Nevertheless, biologists still have major difficulties trying to transfer existing animal- and/or scenario-related machine learning approaches to their specific animal datasets and scientific questions. This study presents an animal-independent, open-source deep learning framework, along with a detailed user guide. Three signal identification tasks, commonly encountered in bioacoustics research, were investigated: (1) target signal vs. background noise detection, (2) species classification, and (3) call type categorization. ANIMAL-SPOT successfully segmented human-annotated target signals in data volumes representing 10 distinct animal species and 1 additional genus, resulting in a mean test accuracy of 97.9%, together with an average area under the ROC curve (AUC) of 95.9%, when predicting on unseen recordings. Moreover, an average segmentation accuracy and F1-score of 95.4% was achieved on the publicly available BirdVox-Full-Night data corpus. In addition, multi-class species and call type classification resulted in 96.6% and 92.7% accuracy on unseen test data, as well as 95.2% and 88.4% regarding previous animal-specific machine-based detection excerpts. Furthermore, an Unweighted Average Recall (UAR) of 89.3% outperformed the multi-species classification baseline system of the ComParE 2021 Primate Sub-Challenge. Besides animal independence, ANIMAL-SPOT does not rely on expert knowledge or special computing resources, thereby making deep-learning-based bioacoustic signal identification accessible to a broad audience.

A Review of Automated Bioacoustics and General Acoustics Classification Research

Automated bioacoustics classification has received increasing attention from the research community in recent years due its cross-disciplinary nature and its diverse application. Applications in bioacoustics classification range from smart acoustic sensor networks that investigate the effects of acoustic vocalizations on species to context-aware edge devices that anticipate changes in their environment adapt their sensing and processing accordingly. The research described here is an in-depth survey of the current state of bioacoustics classification and monitoring. The survey examines bioacoustics classification alongside general acoustics to provide a representative picture of the research landscape. The survey reviewed 124 studies spanning eight years of research. The survey identifies the key application areas in bioacoustics research and the techniques used in audio transformation and feature extraction. The survey also examines the classification algorithms used in bioacoustics systems. Lastly, the survey examines current challenges, possible opportunities, and future directions in bioacoustics.

Computational bioacoustics with deep learning: a review and roadmap

Animal vocalisations and natural soundscapes are fascinating objects of study, and contain valuable evidence about animal behaviours, populations and ecosystems. They are studied in bioacoustics and ecoacoustics, with signal processing and analysis an important component. Computational bioacoustics has accelerated in recent decades due to the growth of affordable digital sound recording devices, and to huge progress in informatics such as big data, signal processing and machine learning. Methods are inherited from the wider field of deep learning, including speech and image processing. However, the tasks, demands and data characteristics are often different from those addressed in speech or music analysis. There remain unsolved problems, and tasks for which evidence is surely present in many acoustic signals, but not yet realised. In this paper I perform a review of the state of the art in deep learning for computational bioacoustics, aiming to clarify key concepts and identify and analyse knowledge gaps. Based on this, I offer a subjective but principled roadmap for computational bioacoustics with deep learning: topics that the community should aim to address, in order to make the most of future developments in AI and informatics, and to use audio data in answering zoological and ecological questions.

Identification, Analysis and Characterization of Base Units of Bird Vocal Communication: The White Spectacled Bulbul (Pycnonotus xanthopygos) as a Case Study

Animal vocal communication is a broad and multi-disciplinary field of research. Studying various aspects of communication can provide key elements for understanding animal behavior, evolution, and cognition. Given the large amount of acoustic data accumulated from automated recorders, for which manual annotation and analysis is impractical, there is a growing need to develop algorithms and automatic methods for analyzing and identifying animal sounds. In this study we developed an automatic detection and analysis system based on audio signal processing algorithms and deep learning that is capable of processing and analyzing large volumes of data without human bias. We selected the White Spectacled Bulbul (Pycnonotus xanthopygos) as our bird model because it has a complex vocal communication system with a large repertoire which is used by both sexes, year-round. It is a common, widespread passerine in Israel, which is relatively easy to locate and record in a broad range of habitats. Like many passerines, the Bulbul’s vocal communication consists of two primary hierarchies of utterances, syllables and words. To extract each of these units’ characteristics, the fundamental frequency contour was modeled using a low degree Legendre polynomial, enabling it to capture the different patterns of variation from different vocalizations, so that each pattern could be effectively expressed using very few coefficients. In addition, a mel-spectrogram was computed for each unit, and several features were extracted both in the time-domain (e.g., zero-crossing rate and energy) and frequency-domain (e.g., spectral centroid and spectral flatness). We applied both linear and non-linear dimensionality reduction algorithms on feature vectors and validated the findings that were obtained manually, namely by listening and examining the spectrograms visually. Using these algorithms, we show that the Bulbul has a complex vocabulary of more than 30 words, that there are multiple syllables that are combined in different words, and that a particular syllable can appear in several words. Using our system, researchers will be able to analyze hundreds of hours of audio recordings, to obtain objective evaluation of repertoires, and to identify different vocal units and distinguish between them, thus gaining a broad perspective on bird vocal communication.

Sounding the Call for a Global Library of Underwater Biological Sounds

Aquatic environments encompass the world’s most extensive habitats, rich with sounds produced by a diversity of animals. Passive acoustic monitoring (PAM) is an increasingly accessible remote sensing technology that uses hydrophones to listen to the underwater world and represents an unprecedented, non-invasive method to monitor underwater environments. This information can assist in the delineation of biologically important areas via detection of sound-producing species or characterization of ecosystem type and condition, inferred from the acoustic properties of the local soundscape. At a time when worldwide biodiversity is in significant decline and underwater soundscapes are being altered as a result of anthropogenic impacts, there is a need to document, quantify, and understand biotic sound sources–potentially before they disappear. A significant step toward these goals is the development of a web-based, open-access platform that provides: (1) a reference library of known and unknown biological sound sources (by integrating and expanding existing libraries around the world); (2) a data repository portal for annotated and unannotated audio recordings of single sources and of soundscapes; (3) a training platform for artificial intelligence algorithms for signal detection and classification; and (4) a citizen science-based application for public users. Although individually, these resources are often met on regional and taxa-specific scales, many are not sustained and, collectively, an enduring global database with an integrated platform has not been realized. We discuss the benefits such a program can provide, previous calls for global data-sharing and reference libraries, and the challenges that need to be overcome to bring together bio- and ecoacousticians, bioinformaticians, propagation experts, web engineers, and signal processing specialists (e.g., artificial intelligence) with the necessary support and funding to build a sustainable and scalable platform that could address the needs of all contributors and stakeholders into the future.

Utilizing DeepSqueak for automatic detection and classification of mammalian vocalizations: a case study on primate vocalizations

Bioacoustic analyses of animal vocalizations are predominantly accomplished through manual scanning, a highly subjective and time-consuming process. Thus, validated automated analyses are needed that are usable for a variety of animal species and easy to handle by non-programing specialists. This study tested and validated whether DeepSqueak, a user-friendly software, developed for rodent ultrasonic vocalizations, can be generalized to automate the detection/segmentation, clustering and classification of high-frequency/ultrasonic vocalizations of a primate species. Our validation procedure showed that the trained detectors for vocalizations of the gray mouse lemur (Microcebus murinus) can deal with different call types, individual variation and different recording quality. Implementing additional filters drastically reduced noise signals (4225 events) and call fragments (637 events), resulting in 91% correct detections (N_total = 3040). Additionally, the detectors could be used to detect the vocalizations of an evolutionary closely related species, the Goodman’s mouse lemur (M. lehilahytsara). An integrated supervised classifier classified 93% of the 2683 calls correctly to the respective call type, and the unsupervised clustering model grouped the calls into clusters matching the published human-made categories. This study shows that DeepSqueak can be successfully utilized to detect, cluster and classify high-frequency/ultrasonic vocalizations of other taxa than rodents, and suggests a validation procedure usable to evaluate further bioacoustics software.

Comparing recurrent convolutional neural networks for large scale bird species classification

We present a deep learning approach towards the large-scale prediction and analysis of bird acoustics from 100 different bird species. We use spectrograms constructed on bird audio recordings from the Cornell Bird Challenge (CBC)2020 dataset, which includes recordings of multiple and potentially overlapping bird vocalizations with background noise. Our experiments show that a hybrid modeling approach that involves a Convolutional Neural Network (CNN) for learning the representation for a slice of the spectrogram, and a Recurrent Neural Network (RNN) for the temporal component to combine across time-points leads to the most accurate model on this dataset. We show results on a spectrum of models ranging from stand-alone CNNs to hybrid models of various types obtained by combining CNNs with other CNNs or RNNs of the following types: Long Short-Term Memory (LSTM) networks, Gated Recurrent Units (GRU), and Legendre Memory Units (LMU). The best performing model achieves an average accuracy of 67% over the 100 different bird species, with the highest accuracy of 90% for the bird species, Red crossbill. We further analyze the learned representations visually and find them to be intuitive, where we find that related bird species are clustered close together. We present a novel way to empirically interpret the representations learned by the LMU-based hybrid model which shows how memory channel patterns change over time with the changes seen in the spectrograms.

Seismic localization of elephant rumbles as a monitoring approach

African elephants (Loxodonta africana) are sentient and intelligent animals that use a variety of vocalizations to greet, warn or communicate with each other. Their low-frequency rumbles propagate through the air as well as through the ground and the physical properties of both media cause differences in frequency filtering and propagation distances of the respective wave. However, it is not well understood how each mode contributes to the animals' abilities to detect these rumbles and extract behavioural or spatial information. In this study, we recorded seismic and co-generated acoustic rumbles in Kenya and compared their potential use to localize the vocalizing animal using the same multi-lateration algorithms. For our experimental set-up, seismic localization has higher accuracy than acoustic, and bimodal localization does not improve results. We conclude that seismic rumbles can be used to remotely monitor and even decipher elephant social interactions, presenting us with a tool for far-reaching, non-intrusive and surprisingly informative wildlife monitoring.

Automatic detection and classification of baleen whale social calls using convolutional neural networks

Passive acoustic monitoring has proven to be an indispensable tool for many aspects of baleen whale research. Manual detection of whale calls on these large data sets demands extensive manual labor. Automated whale call detectors offer a more efficient approach and have been developed for many species and call types. However, calls with a large level of variability such as fin whale (Balaenoptera physalus) 40 Hz call and blue whale (B. musculus) D call have been challenging to detect automatically and hence no practical automated detector exists for these two call types. Using a modular approach consisting of faster region-based convolutional neural network followed by a convolutional neural network, we have created automated detectors for 40 Hz calls and D calls. Both detectors were tested on recordings with high- and low density of calls and, when selecting for detections with high classification scores, they were shown to have precision ranging from 54% to 57% with recall ranging from 72% to 78% for 40 Hz and precision ranging from 62% to 64% with recall ranging from 70 to 73% for D calls. As these two call types are produced by both sexes, using them in long-term studies would remove sex-bias in estimates of temporal presence and movement patterns.

Detecting, classifying, and counting blue whale calls with Siamese neural networks

The goal of this project is to use acoustic signatures to detect, classify, and count the calls of four acoustic populations of blue whales so that, ultimately, the conservation status of each population can be better assessed. We used manual annotations from 350 h of audio recordings from the underwater hydrophones in the Indian Ocean to build a deep learning model to detect, classify, and count the calls from four acoustic song types. The method we used was Siamese neural networks (SNN), a class of neural network architectures that are used to find the similarity of the inputs by comparing their feature vectors, finding that they outperformed the more widely used convolutional neural networks (CNN). Specifically, the SNN outperform a CNN with 2% accuracy improvement in population classification and 1.7%-6.4% accuracy improvement in call count estimation for each blue whale population. In addition, even though we treat the call count estimation problem as a classification task and encode the number of calls in each spectrogram as a categorical variable, SNN surprisingly learned the ordinal relationship among them. SNN are robust and are shown here to be an effective way to automatically mine large acoustic datasets for blue whale calls.

Deep embedded clustering of coral reef bioacoustics

Deep clustering was applied to unlabeled, automatically detected signals in a coral reef soundscape to distinguish fish pulse calls from segments of whale song. Deep embedded clustering (DEC) learned latent features and formed classification clusters using fixed-length power spectrograms of the signals. Handpicked spectral and temporal features were also extracted and clustered with Gaussian mixture models (GMM) and conventional clustering. DEC, GMM, and conventional clustering were tested on simulated datasets of fish pulse calls (fish) and whale song units (whale) with randomized bandwidth, duration, and SNR. Both GMM and DEC achieved high accuracy and identified clusters with fish, whale, and overlapping fish and whale signals. Conventional clustering methods had low accuracy in scenarios with unequal-sized clusters or overlapping signals. Fish and whale signals recorded near Hawaii in February-March 2020 were clustered with DEC, GMM, and conventional clustering. DEC features demonstrated the highest accuracy of 77.5% on a small, manually labeled dataset for classifying signals into fish and whale clusters.

EXACT: a collaboration toolset for algorithm-aided annotation of images with annotation version control

In many research areas, scientific progress is accelerated by multidisciplinary access to image data and their interdisciplinary annotation. However, keeping track of these annotations to ensure a high-quality multi-purpose data set is a challenging and labour intensive task. We developed the open-source online platform EXACT (EXpert Algorithm Collaboration Tool) that enables the collaborative interdisciplinary analysis of images from different domains online and offline. EXACT supports multi-gigapixel medical whole slide images as well as image series with thousands of images. The software utilises a flexible plugin system that can be adapted to diverse applications such as counting mitotic figures with a screening mode, finding false annotations on a novel validation view, or using the latest deep learning image analysis technologies. This is combined with a version control system which makes it possible to keep track of changes in the data sets and, for example, to link the results of deep learning experiments to specific data set versions. EXACT is freely available and has already been successfully applied to a broad range of annotation tasks, including highly diverse applications like deep learning supported cytology scoring, interdisciplinary multi-centre whole slide image tumour annotation, and highly specialised whale sound spectroscopy clustering.

Performance of a deep neural network at detecting North Atlantic right whale upcalls

Passive acoustics provides a powerful tool for monitoring the endangered North Atlantic right whale (Eubalaena glacialis), but robust detection algorithms are needed to handle diverse and variable acoustic conditions and differences in recording techniques and equipment. This paper investigates the potential of deep neural networks (DNNs) for addressing this need. ResNet, an architecture commonly used for image recognition, was trained to recognize the time-frequency representation of the characteristic North Atlantic right whale upcall. The network was trained on several thousand examples recorded at various locations in the Gulf of St. Lawrence in 2018 and 2019, using different equipment and deployment techniques. Used as a detection algorithm on fifty 30-min recordings from the years 2015-2017 containing over one thousand upcalls, the network achieved recalls up to 80% while maintaining a precision of 90%. Importantly, the performance of the network improved as more variance was introduced into the training dataset, whereas the opposite trend was observed using a conventional linear discriminant analysis approach. This study demonstrates that DNNs can be trained to identify North Atlantic right whale upcalls under diverse and variable conditions with a performance that compares favorably to that of existing algorithms.

Beluga whale acoustic signal classification using deep learning neural network models

Over a decade after the Cook Inlet beluga (Delphinapterus leucas) was listed as endangered in 2008, the population has shown no sign of recovery. Lack of ecological knowledge limits the understanding of, and ability to manage, potential threats impeding recovery of this declining population. National Oceanic and Atmospheric Administration Fisheries, in partnership with the Alaska Department of Fish and Game, initiated a passive acoustics monitoring program in 2017 to investigate beluga seasonal occurrence by deploying a series of passive acoustic moorings. Data have been processed with semi-automated tonal detectors followed by time intensive manual validation. To reduce this labor intensive and time-consuming process, in addition to increasing the accuracy of classification results, the authors constructed an ensembled deep learning convolutional neural network model to classify beluga detections as true or false. Using a 0.5 threshold, the final model achieves 96.57% precision and 92.26% recall on testing dataset. This methodology proves to be successful at classifying beluga signals, and the framework can be easily generalized to other acoustic classification problems.

ORCA-SPOT: An Automatic Killer Whale Sound Detection Toolkit Using Deep Learning

Large bioacoustic archives of wild animals are an important source to identify reappearing communication patterns, which can then be related to recurring behavioral patterns to advance the current understanding of intra-specific communication of non-human animals. A main challenge remains that most large-scale bioacoustic archives contain only a small percentage of animal vocalizations and a large amount of environmental noise, which makes it extremely difficult to manually retrieve sufficient vocalizations for further analysis - particularly important for species with advanced social systems and complex vocalizations. In this study deep neural networks were trained on 11,509 killer whale (Orcinus orca) signals and 34,848 noise segments. The resulting toolkit ORCA-SPOT was tested on a large-scale bioacoustic repository - the Orchive - comprising roughly 19,000 hours of killer whale underwater recordings. An automated segmentation of the entire Orchive recordings (about 2.2 years) took approximately 8 days. It achieved a time-based precision or positive-predictive-value (PPV) of 93.2% and an area-under-the-curve (AUC) of 0.9523. This approach enables an automated annotation procedure of large bioacoustics databases to extract killer whale sounds, which are essential for subsequent identification of significant communication patterns. The code will be publicly available in October 2019 to support the application of deep learning to bioaoucstic research. ORCA-SPOT can be adapted to other animal species.