Whistle Contours Research Articles

Extracting dolphin whistles in complex acoustic scenarios: a case study in the Bay of Biscay

ABSTRACT Accurate whistle contour extraction is crucial in many dolphin behavioural studies. Traditionally, whistle contour extraction involves a first step of finding whistle candidates by peak-level detection in the time-frequency domain, followed by a determination of when peaks are close enough to each other to be part of the same whistle contour. In complex scenarios, such as those with a large number of individuals vocalising simultaneously or those with a sudden increase in background noise, peak-level detection may not provide a number of accurate whistle candidates that is large enough to extract the whistle contour or to disambiguate individual whistles when they cross one another. In these adverse scenarios, a different approach, based on the pyknogram representation, can produce a more accurate detection of whistle candidates and evenly distributed candidates throughout the duration of the whistle. This work compares the peak-level extraction approach of the spectrogram with the point-density extraction approach of the pyknogram. We propose a technique that combines estimates of the central frequency and bandwidth to extract whistle candidates in adverse scenarios. The method has been successfully used for the vocalisation extraction of dolphins in the Bay of Biscay (Spain) using a database of more than 2000 dolphin whistles.

First identification of stereotyped whistle contour types by Pseudorca crassidens

First identification of stereotyped whistle contour types by <i>Pseudorca crassidens</i>

Signature whistle use and changes in whistle emission rate in a rehabilitated rough-toothed dolphin

Acoustic signals play a crucial role in communication among animals, particularly in dolphins. Signature whistles, one of their most extensively studied vocalizations, enable dolphins to convey their identity to conspecifics through individually distinct whistle contours. However, it remains unclear whether rough-toothed dolphins (Steno bredanensis) also produce signature whistles with individually identifying contours and, if so, whether they are associated with stress and poor health, such as in bottlenose dolphins. To bridge this knowledge gap, we recorded sounds emitted by a live-stranded rough-toothed dolphin during its rehabilitation in May 2017 at Isla Mujeres, Quintana Roo, Mexico. We assessed if the dolphin produced a signature whistle and whether whistle rate, inter-whistle interval, mean low and high frequencies, and blood chemistry measures, changed significantly over time. While isolated from conspecifics during rehabilitation, the dolphin generated a single, repeated, and stereotyped whistle contour that met the previously established SIGnature IDentification criteria for signature whistle emissions for bottlenose dolphins. Whistle characteristics varied over the 11 recording days: whistle rate and inter-whistle interval significantly decreased over time; the number of whistles with preceding echolocation click trains decreased over time; and mean low and high frequencies changed over recording days. We conclude that this rough-toothed dolphin possessed what resembles a signature whistle contour, and the emission of this contour underwent significant changes throughout the rehabilitation process. While our study presents evidence of a single rough-toothed dolphin producing a signature whistle, further research is necessary to determine whether this vocal behavior is prevalent across the species.

Cetaceans of the Saya de Malha bank region, Indian Ocean: A candidate Important Marine Mammal Area

The banks of Saya de Malha and the surrounding Mascarene Plateau in the Indian Ocean are among the least studied shallow sea water regions in the world. The steep sloping bathymetry of this region is thought to drive high levels of primary productivity and support a diverse range of marine species. Until now, no surveys for cetaceans have been conducted in the waters surrounding Saya de Malha, and so the diversity of cetaceans is unknown. Opportunistic visual and acoustic surveys for cetaceans were conducted onboard a platform of opportunity, the MY Arctic Sunrise, which was involved in a wider project to document the marine life in the region. Survey effort was conducted over 7,700 km, with twelve species of cetacean encountered, including Bryde’s, sperm, beaked and killer whales, along with spinner, striped, pantropical spotted and bottlenose dolphins. A match of a sperm whale coda vocalisation to another, well-studied population off the coast of Mauritius suggests possible connectivity between these regions although further data would be required to confirm this. The banks of Saya de Malha appear to support a diverse range of cetacean species and further systematic surveys are required to increase our understanding of how different species utilise the banks. We provide whistle contours from visually confirmed acoustic detections to contribute towards building a region-specific whistle classifier. Given the diversity of species detected in the region, we suggest that the Saya de Malha bank area be designated, either as an Important Marine Mammal Area or as a marine protected area.

Evidence of signature whistles produced by Indian Ocean bottlenose dolphins (Tursiops aduncus) in Mozambique

ABSTRACT Individually distinctive signature whistles are used by common bottlenose dolphins (Tursiops truncatus) during social interactions and to facilitate group cohesion. There is evidence from a few populations that Indian Ocean bottlenose dolphins (T. aduncus) also use signature whistles. We investigated this from a coastal resident population of T. aduncus in Mozambique. Video-audio data were collected during recreational swim-with activities over 12 years, where potential signature whistles were identified from 28 individuals. Of the 405 whistles documented, 75% were produced in SIGID bouts. Visual assessment of individual signature whistle contours demonstrated temporal stability for up to 8 years. Overall, most signature whistle types had upsweep frequency modulation and were emitted as multiloop whistles. Comparing all whistle contours to each other indicated low discrimination between individuals, with contours frequently categorised together. However, sex differences in the frequency characteristics of whistles were identified, with females whistling at lower frequencies than males. Our results indicate either a shared whistle repertoire or identity encoding with subtle contour features, requiring acute auditory perception and discrimination to decipher caller identity. More widespread geographic investigation into signature whistle use may demonstrate variation in acoustic communication systems for bottlenose dolphins, which are thus far not well understood.

Using deep learning to track time × frequency whistle contours of toothed whales without human-annotated training data.

Many odontocetes produce whistles that feature characteristic contour shapes in spectrogram representations of their calls. Automatically extracting the time × frequency tracks of whistle contours has numerous subsequent applications, including species classification, identification, and density estimation. Deep-learning-based methods, which train models using analyst-annotated whistles, offer a promising way to reliably extract whistle contours. However, the application of such methods can be limited by the significant amount of time and labor required for analyst annotation. To overcome this challenge, a technique that learns from automatically generated pseudo-labels has been developed. These annotations are less accurate than those generated by human analysts but more cost-effective to generate. It is shown that standard training methods do not learn effective models from these pseudo-labels. An improved loss function designed to compensate for pseudo-label error that significantly increases whistle extraction performance is introduced. The experiments show that the developed technique performs well when trained with pseudo-labels generated by two different algorithms. Models trained with the generated pseudo-labels can extract whistles with an F1-score (the harmonic mean of precision and recall) of 86.31% and 87.2% for the two sets of pseudo-labels that are considered. This performance is competitive with a model trained with 12 539 expert-annotated whistles (F1-score of 87.47%).

Differences in the whistles of two ecotypes of bottlenose dolphins from the Gulf of California.

Intraspecific whistle variation in dolphins has been explained in the context of geographical isolation, soundscape, and social structure. Whistles of two ecotypes of bottlenose dolphins from La Paz Bay, Gulf of California, were analyzed. Both ecotypes showed similar whistle contours. However, they could be identified by contour maximum frequency which was predominantly above 15 kHz for oceanic dolphins and below 15 kHz for coastal dolphins. The different whistle frequencies of the two ecotypes could result from differences in group size and acoustic characteristics of their respective habitats, suggesting potential future passive acoustic monitoring applications.

Initial results on the variation of whistle characteristics of bottlenose dolphins from two neighbouring regions of the Mediterranean Sea: Northern Ionian and Southern Adriatic Sea

Bottlenose dolphins have a complex vocal repertoire that varies depending on behavioral context, social structure, group composition, and anthropogenic pressures. This current study describes the whistle characteristics of bottlenose dolphins for the first time in the South Adriatic Sea while assessing the potential differences between whistle characteristics of geographically separated dolphins within neighbouring waters of the North Ionian Sea. The results show that whistle characteristics were similar between Taranto Gulf (Italy) and Boka Bay (Montenegro), despite their spatial differences. The mean peak frequency was 10kHz for each study location while the mean minimum and maximum frequency ranged from 7 to 14kHz. The average duration of whistles was 500 milliseconds. These results share similarities with previous literature, although several studies reported slightly different mean peak frequencies, ranging up to 15kHz in the neighbouring waters of Croatia and Italy. Further, harmonics were produced and formed in 40% of the whistles in Taranto Gulf and 30% of the whistles in Boka Bay. A high incidence of harmonics has previously been associated with behavioral states (i.e., travelling) and with certain types of marine traffic (i.e., fishing vessels). Therefore, it is important to collect simultaneous data on the visual behavior of the focal group as well as document the type and density of marine traffic within the proximity of the dolphins to have an in-depth understanding of vocal behavior. Despite the similarities of whistle characteristics of Taranto and Boka Bay, the whistle contours showed notable variations. Upsweep whistles were the most regularly produced whistle type in each location, which coincides with previous studies in the Mediterranean Sea. However, the least produced whistle had a concave contour in Taranto and was flat in Boka Bay. Previous studies have confirmed that flat whistles account for the least produced whistle contour in the Mediterranean Basin. Examining the whistle characteristics and the variation in whistle contours provides an in-depth understanding of the behavioral complexity as well as its plasticity in the presence of pressure. Therefore, future studies need to include behavior, group composition, noise levels, and human presence to enable an effective understanding of variation in whistle characteristics of bottlenose dolphins.

Learning Stage-Wise GANs for Whistle Extraction in Time-Frequency Spectrograms

Whistle contour extraction aims to derive animal whistles from time-frequency spectrograms as polylines. For toothed whales, whistle extraction results can serve as the basis for analyzing animal abundance, species identity, and social activities. During the last few decades, as long-term recording systems have become affordable, automated whistle extraction algorithms were proposed to process large volumes of recording data. Recently, a deep learning-based method demonstrated superior performance in extracting whistles under varying noise conditions. However, training such networks requires a large amount of labor-intensive annotation, which is not available for many species. To overcome this limitation, we present a framework of stage-wise generative adversarial networks (GANs), which compile new whistle data suitable for deep model training via three stages: generation of background noise in the spectrogram, generation of whistle contours, and generation of whistle signals. By separating the generation of different components in the samples, our framework composes visually promising whistle data and labels even when few expert annotated data are available. Regardless of the amount of human-annotated data, the proposed data augmentation framework leads to a consistent improvement in performance of the whistle extraction model, with a maximum increase of 1.69 in the whistle extraction mean F1-score. Our stage-wise GAN also surpasses one single GAN in improving whistle extraction models with augmented data. The data and code will be available at https://github.com/Paul-LiPu/CompositeGAN\_WhistleAugment.

Probable signature whistle production in Atlantic white‐sided (Lagenorhynchus acutus) and short‐beaked common (Delphinus delphis) dolphins near Cape Cod, Massachusetts

Probable signature whistle production in Atlantic white‐sided (<i>Lagenorhynchus acutus</i>) and short‐beaked common (<i>Delphinus delphis</i>) dolphins near Cape Cod, Massachusetts

Killer Whale (Orcinus orca) Pulsed Calls in the Eastern Canadian Arctic

&#x0D; Killer whales (Orcinus orca) produce a variety of acoustic signal types used for communication: clicks, whistles, and pulsed calls. Discrete pulsed calls are highly stereotyped, repetitive, and unique to individual pods found around the world. Discriminating amongst pod specific calls can help determine population structure in killer whales and is used to track pod movements around oceans. Killer whale presence in the Canadian Arctic has increased substantially, but we have limited understanding of their ecology, movements, and stock identity. Two autonomous passive acoustic monitoring (PAM) hydrophones were deployed in the waters of Eclipse Sound and Milne Inlet, in northern Baffin Island, Nunavut, Canada, in August and September 2017. Eleven killer whale pulsed call types, three multiphonic and eight monophonic, are proposed and described using manual whistle contour extraction and feature normalization. Automated detection of echolocation clicks between 20 and 48 kHz demonstrated little to no overlap between killer whale calls and echolocation presumed to be narwhal, which suggests that narwhal remain audibly inconspicuous when killer whales are present. Describing the acoustic repertoire of killer whales seasonally present in the Canadian Arctic will aid in understanding their acoustic behaviour, seasonal movements, and ecological impacts. The calls described here provide a basis for future acoustic comparisons across the North Atlantic and aid in characterizing killer whale demographics and ecology, particularly for pods making seasonal incursions into Arctic waters.

Semantic segmentation-based whistle extraction of Indo-Pacific Bottlenose Dolphin residing at the coast of Jeju island

Passive acoustic monitoring (PAM) is commonly utilized to monitor cetacean species’ distribution, abundance, and behavior. The demand for automated methods to detect and extract cetacean vocalizations from acoustic data has increased in the last few decades. Automatic whistle extraction of Indo-Pacific Bottlenose Dolphin (IPBD) and other whistle-producing delphinids habitating in the coastal areas represents a challenging problem due to the high ambient noise, including ship noise snapping shrimp at the same habitat. The acoustic signal containing snapping shrimp sound was usually excluded during the development of the detection method. A robust tool of bioacoustics for snapping shrimp sound is still lacking. This study trained a convolutional neural network (CNN) designed for semantic segmentation, initially developed for autonomous driving, to extract the whistle contour from spectrogram at a pixel level. A total of 1600 datasets was annotated for training and testing. As a result, the semantic segmentation classified the whistle with an overall mean Precision of 0.96, Accuracy of 0.89, and F-score of 0.86. In particular, the semantic segmentation extracted the whistle even if associated with a rich snapping shrimp sound, which the conventional method is incapable of. The advancement of metrics presented in this paper will enable long-term assessment of the IPBD population and individual or group tracking.

Contour variations and acoustic characteristics of whistle sounds emitted by Pacific white‐sided dolphins in shallow coastal waters of the Sea of Japan

AbstractPacific white‐sided dolphins, Lagenorhynchus obliquidens (PWSDs), migrate along the coast of Japan. PWSDs rarely emit whistles, instead utilizing pulses, unlike many other delphinid species that frequently emit whistle sounds. In the present study, we aimed to clarify the acoustic characteristics and repertoire of the whistle of PWSDs, which have not been reported so far. We analyzed PWSD whistle sound systems (e.g., acoustic parameters and contours) during boat surveys and using passive acoustic monitoring in shallow coastal waters in the northern Sea of Japan. PWSDs in Shakotan rarely emitted whistles, while the burst pulse frequently was as reported in previous studies in other waters. The PWSD whistle vocalization rates were higher during traveling than during foraging. The frequency and duration of their whistles did not appear to change between behaviors. Although complex multiloop whistles were not recorded, whistle contours were classified into seven types. Whistles are often recorded intensively in a short period as a sequence or bout. Considering the extremely low whistle vocalization rates, dolphins may choose whistles instead of pulsed sounds in limited contexts. Our results shed light on the differences in whistle sound usage within odontocete species.

Non-stereotyped amplitude modulation across signature whistle contours

Bottlenose dolphin signature whistles are characterized by distinctive frequency modulation over time. The stable frequency contours of these whistles broadcast individual identity information. Little is known however, about whether or not the amplitude contour is also stereotyped. Here, we examined the relative amplitude-time contour of signature whistle emissions from eight bottlenose dolphins (Tursiops truncatus) in the U.S. Navy Marine Mammal Program (MMP) in San Diego, CA. The results suggested that unlike the stable frequency-time contour, the amplitude-time contour of signature whistles were largely non-stereotyped, characterized by large variability across multiple whistle emissions. Relative amplitude was negatively related to log peak frequency, with more energy focused in the lower frequency bands. This trend was consistent over all eight dolphins despite having quite different signature whistle contours. This relationship led to the amplitude contours being slightly more stereotyped within than between dolphins. We propose that amplitude across signature whistle emissions may serve as an avenue for encoding additional communicative information. We encourage future studies to incorporate analyses of amplitude contours in addition to frequency contours of signature whistles in order to begin to understand what role it may play in the dolphin communication system.

A comparison of transient killer whale (Orcinus orca) whistles in Alaska

Several species of dolphins in both isolated and continuous regions exhibit geographic variation in whistles. Although whistles are thought to be learned, possible structural components are exhibited across communities. Whistles from three transient killer whale (Orcinus Orca) groups, Prince William Sound, the Pribilof Islands, and False Pass were compared for spectral and temporal properties. Of 395 randomly selected frequency contours of whistles, there were 32 shape categories of whistles. Overall, whistle shapes from different areas reflect homogeneity, with four whistle shapes comprising 50% of all repertoires. The most common whistle contour shapes were up, down, flat, and oscillatory. Whistles from at St. Paul Island, described here for the first time, had higher fundamental frequencies and shorter durations than whistles from Prince William Sound and False Pass. Surface whistles from St. Paul Island had higher mean frequencies than those recorded at depth with passive acoustics. One identical whistle type was found at both St. Paul Island and False Pass, which suggests the same whales travel between the two areas. The longest whistles and most variable whistles were recorded at False Pass.

Species information in whistle frequency modulation patterns of common dolphins.

The most flexible communication systems are those of open-ended vocal learners that can acquire new signals throughout their lifetimes. While acoustic signals carry information in general voice features that affect all of an individual's vocalizations, vocal learners can also introduce novel call types to their repertoires. Delphinids are known for using such learned call types in individual recognition, but their role in other contexts is less clear. We investigated the whistles of two closely related, sympatric common dolphin species, Delphinus delphis and Delphinus bairdii, to evaluate species differences in whistle contours. Acoustic recordings of single-species groups were obtained from the Southern California Bight. We used an unsupervised neural network to categorize whistles and compared the resulting whistle types between species. Of the whistle types recorded in more than one encounter, 169 were shared between species and 60 were species-specific (32 D. delphis types, 28 D. bairdii types). Delphinus delphis used 15 whistle types with an oscillatory frequency contour while only one such type was found in D. bairdii. Given the role of vocal learning in delphinid vocalizations, we argue that these differences in whistle production are probably culturally driven and could help facilitate species recognition between Delphinus species.This article is part of the theme issue ‘Vocal learning in animals and humans’.

Vocal learning in animals and humans.

Vocal learning in animals and humans.

Whistles of the pantropical spotted dolphin (Stenella attenuata) in Santos Basin, western South Atlantic Ocean.

The vocal repertoire of the pantropical spotted dolphin (Stenella attenuata) is poorly documented, with no published information about acoustic signals from South Atlantic Ocean populations. We conducted passive acoustic monitoring and recording of S. attenuata population in the Santos Basin, Brazil, using a towed hydrophone array during line-transects surveys. Our monitoring yielded whistle samples derived from eight groups of S. attenuata, from which we selected 155 whistles for further analysis. Approximately 48% of the whistles presented ultrasonic frequency values, with maximum frequencies up to 31.1 kHz. Across the sample, the number of steps ranged from 0 to 20 and inflection points ranged from 0 to 8. On average, end frequencies were higher than start frequencies, and whistles generally presented wide frequency ranges, with an average of 11.3 kHz. The most predominant whistle contour category was "ascending-descending." Our study provides new information regarding the acoustic repertoire of this poorly documented species and will aid efforts for using acoustics to identify and monitor cetaceans in this region.

Oscillatory whistles—The ups and downs of identifying species in passive acoustic recordings

Sympatric short- and long-beaked common dolphins in the Southern California Bight (Delphinus delphis and D. delphis bairdii) are challenging to identify acoustically because their whistles overlap in many time-frequency characteristics. We therefore asked whether frequency modulation patterns can help with species identification. Whistle contours from single-species encounters (short-beaked = 902 whistles, 14 schools, long-beaked = 872 whistles, 10 schools) were extracted and categorized based on frequency content and shape using dynamic time warping and artificial neural networks. This analysis resulted in 447 whistle types with 38% being produced by both species. Of the remaining species-specific whistle types, 22% (n = 60) were recorded from more than one school. Thirty-two of these were specific to short-beaked common dolphins and 28 were specific to long-beaked common dolphins. Almost half of the short-beaked common dolphin species-specific whistle types (47%) were oscillatory (contour shape containing at least two cycles with the maximum and minimum of each cycle separated by at least 1 kHz), while only 3% of long-beaked common dolphin species-specific types were oscillatory. Thus, oscillatory whistles appear to be diagnostic of short-beaked common dolphins in this area. More broadly, our findings suggest that repertoire-wide comparisons of acoustic features may overlook possible species recognition via specific signals.

A frequency hopping pattern inspired bionic underwater acoustic communication

The covert underwater acoustic communication realized from the perspective of bionics draws an increasing interest in recent researches. Inspired by the frequency hopping spectrogram pattern, a bionic underwater acoustic communication method is proposed by using cetacean whistle as communication carrier. The basic idea is to generate synthetic whistle with different spectrogram to transmit the information. The whistle is first modeled after extracting the whistle contour from the noise-polluted recordings. Then a synthetic whistle with a special spectrogram pattern which mimics the whistle spectral characteristic is generated to modulate the information. The bionic communication frame is the combination of the original whistle and the synthetic whistles which are used for synchronization and information modulation respectively. After synchronization, the each received symbol is extracted and correlated with the locally generated signals. The information demodulation is conducted by the correlation peak decision. Finally the bionic effect of communication signals is analyzed through time-frequency correlation coefficient and Mel frequency cepstrum coefficient. Simulation and sea trial verified the feasibility of the proposed method. Less than 10−3 uncoded bit error rate is achieved with the correlation coefficient over 0.9 at a distance of 8.5km in the sea trial.