A real-time method for autonomous passive acoustic detection-classification of humpback whales

This paper describes a method for real-time, autonomous, joint detection-classification of humpback whale vocalizations. The approach adapts the spectrogram correlation method used by Mellinger and Clark [J. Acoust. Soc. Am. 107, 3518-3529 (2000)] for bowhead whale endnote detection to the humpback whale problem. The objective is the implementation of a system to determine the presence or absence of humpback whales with passive acoustic methods and to perform this classification with low false alarm rate in real time. Multiple correlation kernels are used due to the diversity of humpback song. The approach also takes advantage of the fact that humpbacks tend to vocalize repeatedly for extended periods of time, and identification is declared only when multiple song units are detected within a fixed time interval. Humpback whale vocalizations from Alaska, Hawaii, and Stellwagen Bank were used to train the algorithm. It was then tested on independent data obtained off Kaena Point, Hawaii in February and March of 2009. Results show that the algorithm successfully classified humpback whales autonomously in real time, with a measured probability of correct classification in excess of 74% and a measured probability of false alarm below 1%.     

Low frequency vocalizations attributed to sei whales (Balaenoptera borealis)

Low frequency (<100 Hz) downsweep vocalizations were repeatedly recorded from ocean gliders east of Cape Cod, MA in May 2005. To identify the species responsible for this call, arrays of acoustic recorders were deployed in this same area during 2006 and 2007. 70 h of collocated visual observations at the center of each array were used to compare the localized occurrence of this call to the occurrence of three baleen whale species: right, humpback, and sei whales. The low frequency call was significantly associated only with the occurrence of sei whales. On average, the call swept from 82 to 34 Hz over 1.4 s and was most often produced as a single call, although pairs and (more rarely) triplets were occasionally detected. Individual calls comprising the pairs were localized to within tens of meters of one another and were more similar to one another than to contemporaneous calls by other whales, suggesting that paired calls may be produced by the same animal. A synthetic kernel was developed to facilitate automatic detection of this call using spectrogram-correlation methods. The optimal kernel missed 14% of calls, and of all the calls that were automatically detected, 15% were false positives.     

Sounds produced by Norwegian killer whales, Orcinus orca, during capture

To date very little is still known about the acoustic behavior of Norwegian killer whales, in particular that of individual whales. In this study a unique opportunity was presented to document the sounds produced by five captured killer whales in the Vestfjord area, northern Norway. Individuals produced 14 discrete and 7 compound calls. Two call types were used both by individuals 16178 and 23365 suggesting that they may belong to the same pod. Comparisons with calls documented in Strager (1993) showed that none of the call types used by the captured individuals were present. The lack of these calls in the available literature suggests that call variability within individuals is likely to be large. This short note adds to our knowledge of the vocal repertoire of this population and demonstrates the need for further studies to provide behavioural context to these sounds.     

20-Hz pulses and other vocalizations of fin whales, Balaenoptera physalus, in the Gulf of California, Mexico.

Low-frequency vocalizations were recorded from fin whales, Balaenoptera physalus, in the Gulf of California, Mexico, during three cruises. In March 1985, recorded 20-Hz pulses were in sequences of regular 9-s interpulse intervals. In August 1987, nearly all were in sequences of doublets with alternating 5- and 18-s interpulse intervals. No 20-Hz pulse sequences of any kind were detected in February 1987. The typical pulse modulated from 42 to 20 Hz and its median duration was 0.7 s (1985 data). Most other fin whale sounds were also short tonal pulses averaging 82, 56, and 68 Hz, respectively, for the three cruises; 89% were modulated in frequency, mostly downward. Compared to Atlantic and Pacific Ocean regions, Gulf of California 20-Hz pulses were unique in terms of frequency modulation, interpulse sound levels, and temporal patterns. Fin whales in the Gulf may represent a regional stock revealed by their sound characteristics, a phenomenon previously shown for humpback whales, birds, and fish. Regional differences in fin whale sounds were found in comparisons of Atlantic and Pacific locations.     

Subunit definition and analysis for humpback whale call classification

Songs of humpback whales (Megaptera novaeangliae) have been studied for several years to gain a deeper insight on the intraspecific social interactions. Such a complex acoustic display is indeed thought to play an important role in both the mating ritual and male to male interaction. Hence, the need to classify the unit constituents of a song objectively and systematically has become crucial to allow processing large data sets. We propose a new approach for song segmentation based on the definition of subunits. Songs of humpback whales collected in Madagascar in August 2008 and 2009 were segmented using an energy detector with a double threshold and classified automatically with a clustering algorithm using MFCCs: the results, which were checked against a manual classification, showed that the use of subunit as the basic constituent of a song rather than the unit produces a more accurate classification of the calls. Such results were expected given that subunits are generally shorter in duration and less variable in terms of their frequency content and so their characteristics are more easily captured by an automatic classifier. Analysis of songs from other years and various areas of the World is necessary to corroborate the repeatability of the method proposed.     

Sound production by North Atlantic right whales (Eubalaena glacialis) in surface active groups

The surface active group (SAG) is the most commonly observed surface social behavior of North Atlantic right whales. Recordings were made from 52 SAGs in the Bay of Fundy, Canada between July and September, from 1999 to 2002. The call types recorded from these groups were similar to those described previously for Southern right whales (Eubalaena australis), with six major call types being termed scream, gunshot, blow, upcall, warble, and downcall. The percentage of total calls of each call type depended on the group size and composition. The most common call type recorded was the scream call. The scream calls were produced by the focal female in a SAG. Production of other sound types can be attributed to whales other than the focal female, with gunshot and upcalls produced by males, and warble calls produced by female calves. The source levels for these sounds range from 137 to 162 dB rms re 1 ,tPa-m for tonal calls and 174 to 192 dB rms for broadband gunshot sounds.     

Acoustic analysis of vocal development in a New World primate, the common marmoset (Callithrix jacchus)

In contrast to humans and songbirds, there is limited evidence of vocal learning in nonhuman primates. While previous studies suggested that primate vocalizations exhibit developmental changes, detailed analyses of the extent and time course of such changes across a species' vocal repertoire remain limited. In a highly vocal primate, the common marmoset (Callithrix jacchus), we studied developmental changes in the acoustic structure of species-specific communication sounds produced in a social setting. We performed detailed acoustic analyses of the spectral and temporal characteristics of marmoset vocalizations during development, comparing differences between genders and twin pairs, as well as with vocalizations from adult marmosets residing in the same colony. Our analyses revealed significant changes in spectral and temporal features as well as variability of particular call types over time. Infant and juvenile vocalizations changed progressively toward the vocalizations produced by adult marmosets. Call types observed early in development that were unique to infants disappeared gradually with age, while vocal exchanges with conspecifics emerged. Our observations clearly indicate that marmoset vocalizations undergo both qualitative and quantitative postnatal changes, establishing the basis for further studies to delineate contributions from maturation of the vocal apparatus and behavioral experience.     

Analysis of pain-related vocalization in young pigs

The assessment of pain constitutes a major issue for animal welfare research. The objective of this study was to classify vocalizations during castration pain and to assess alterations in vocalizations under local anaesthesia. The alterations in vocalization were measured by multiparametric call analysis. A total of 4537 calls of 70 young pigs were evaluated. With the data of this study three call types are distinguishable (grunt, squeal, scream). A high percentage (94.64%) of calls that could be classified in one of the three call types during the castration process within the confidence level of 95% was found. The comparison of the occurrence of the call types during treatments gives evidence for pain-related use of screams. The piglets castrated without local anaesthesia produced almost double the number of screams as piglets castrated with anaesthesia. The comparison of the recorded sound parameters reveals the particular position of screams in the call repertoire of young pigs. Screams are significantly different in their sound parameters than grunts or squeals. Castration in comparison to mere restraint produced a comprehensive change in sound parameters, with castration calls becoming more extended and more powerful. The findings in this study also show differences in the effectiveness of the parameters which indicate pain. Parameters that describe a single event in a call, such as peak level or peak frequency give better results than parameters that describe an average, such as weighted frequency and main frequency. The research indicated that pain-related changes of calls in piglets can be identified. On the basis of the results, automatic classification of call types during management operations may be developed. This could contribute to objective animal welfare assessment.     

Environmental constraints on sound transmission by humpback whales

Singing humpback whales in Hawaii produce a variety of sounds at high source levels (ca. 185 dB re: 1 microPa), in coastal waters 15-500 m deep. These sounds are attenuated and distorted as they propagate away from a singer, limiting the utilizable range of the sounds. In the current study, simulations based on normal-mode theory were used to investigate how the effects of shallow-water propagation constrain humpback whales' use of sound. It is shown that humpbacks can greatly affect transmission range by adjusting their positions and sounds in response to environmental factors. Source depth, in particular, is shown to be a major determinant of which frequencies propagate the farthest. A preliminary analysis of range-dependent distortion suggests that spectral cues can potentially provide listening whales with information about how far a sound has traveled.     

Effects of noise levels and call types on the source levels of killer whale calls

Accurate parameter estimates relevant to the vocal behavior of marine mammals are needed to assess potential effects of anthropogenic sound exposure including how masking noise reduces the active space of sounds used for communication. Information about how these animals modify their vocal behavior in response to noise exposure is also needed for such assessment. Prior studies have reported variations in the source levels of killer whale sounds, and a more recent study reported that killer whales compensate for vessel masking noise by increasing their call amplitude. The objectives of the current study were to investigate the source levels of a variety of call types in southern resident killer whales while also considering background noise level as a likely factor related to call source level variability. The source levels of 763 discrete calls along with corresponding background noise were measured over three summer field seasons in the waters surrounding the San Juan Islands, WA. Both noise level and call type were significant factors on call source levels (1-40 kHz band, range of 135.0-175.7 dB(rms) re 1 [micro sign]Pa at 1 m). These factors should be considered in models that predict how anthropogenic masking noise reduces vocal communication space in marine mammals.     

Classification of vocalizations in the Mongolian gerbil, Meriones unguiculatus

The Mongolian gerbil (Meriones unguiculatus) has been an important model system in auditory physiology, but its natural sounds are not well known. Vocalizations produced by colonies of adult gerbils were recorded during various social interactions in a standard laboratory animal-rearing facility. Sound recordings were made continuously for 24 h. This species exhibited a rich repertoire of vocalizations that varied in spectrotemporal structure. Calls were classified into 13 distinct syllable types. These syllables were further categorized into eight simple syllables and five composite syllables, which could be described by combinations of two to three simple syllables. The durations of individual syllables ranged from 30 to 330 ms with fundamental frequencies of 5 to 50 kHz. Those with lower fundamental frequencies typically contained more harmonic components (up to nine). Analysis of syllable sequences indicated that syllables may be combined into three types of simple phrases. These results provide a basis for future studies not only of the behavioral significance of vocalization, but also of the neural basis of vocal communication in the Mongolian gerbil.     

Divergence of a stereotyped call in northern resident killer whales

Northern resident killer whale pods (Orcinus orca) have distinctive stereotyped pulsed call repertoires that can be used to distinguish groups acoustically. Repertoires are generally stable, with the same call types comprising the repertoire of a given pod over a period of years to decades. Previous studies have shown that some discrete pulsed calls can be subdivided into variants or subtypes. This study suggests that new stereotyped calls may result from the gradual modification of existing call types through subtypes. Vocalizations of individuals and small groups of killer whales were collected using a bottom-mounted hydrophone array in Johnstone Strait, British Columbia in 2006 and 2007. Discriminant analysis of slope variations of a predominant call type, N4, revealed the presence of four distinct call subtypes. Similar to previous studies, there was a divergence of the N4 call between members of different matrilines of the same pod. However, this study reveals that individual killer whales produced multiple subtypes of the N4 call, indicating that divergence in the N4 call is not the result of individual differences, but rather may indicate the gradual evolution of a new stereotyped call.     

Acoustic properties of humpback whale songs

A vertical array of five hydrophones was used to measure the acoustic field in the vertical plane of singing humpback whales. Once a singer was located, two swimmers with snorkel gear were deployed to determine the orientation of the whale and position the boat so that the array could be deployed in front of the whale at a minimum standoff distance of at least 10 m. The spacing of the hydrophones was 7 m with the deepest hydrophone deployed at a depth of 35 m. An eight-channel TASCAM recorder with a bandwidth of 24 kHz was used to record the hydrophone signals. The location (distance and depth) of the singer was determined by computing the time of arrival differences between the hydrophone signals. The maximum source level varied between individual units in a song, with values between 151 and 173 dB re 1 microPa. One of the purposes of this study was to estimate potential sound exposure of nearby conspecifics. The acoustic field determined by considering the relative intensity of higher frequency harmonics in the signals indicated that the sounds are projected in the horizontal direction despite the singer being canted head downward anywhere from about 25 degrees to 90 degrees. High-frequency harmonics extended beyond 24 kHz, suggesting that humpback whales may have an upper frequency limit of hearing as high as 24 kHz.     

Cepstral coefficients and hidden Markov models reveal idiosyncratic voice characteristics in red deer (Cervus elaphus) stags

Bouts of vocalizations given by seven red deer stags were recorded over the rutting period, and homomorphic analysis and hidden Markov models (two techniques typically used for the automatic recognition of human speech utterances) were used to investigate whether the spectral envelope of the calls was individually distinctive. Bouts of common roars (the most common call type) were highly individually distinctive, with an average recognition percentage of 93.5%. A "temporal" split-sample approach indicated that although in most individuals these identity cues held over the rutting period, the ability of the models trained with the bouts of roars recorded early in the rut to correctly classify later vocalizations decreased as the recording date increased. When Markov models trained using the bouts of common roars were used to classify other call types according to their individual membership, the classification results indicated that the cues to identity contained in the common roars were also present in the other call types. This is the first demonstration in mammals other than primates that individuals have vocal cues to identity that are common to the different call types that compose their vocal repertoire.     

Low-frequency whale and seismic airgun sounds recorded in the mid-Atlantic Ocean

Beginning in February 1999, an array of six autonomous hydrophones was moored near the Mid-Atlantic Ridge (35 degrees N-15 degrees N, 50 degrees W-33 degrees W). Two years of data were reviewed for whale vocalizations by visually examining spectrograms. Four distinct sounds were detected that are believed to be of biological origin: (1) a two-part low-frequency moan at roughly 18 Hz lasting 25 s which has previously been attributed to blue whales (Balaenoptera musculus); (2) series of short pulses approximately 18 s apart centered at 22 Hz, which are likely produced by fin whales (B. physalus); (3) series of short, pulsive sounds at 30 Hz and above and approximately 1 s apart that resemble sounds attributed to minke whales (B. acutorostrata); and (4) downswept, pulsive sounds above 30 Hz that are likely from baleen whales. Vocalizations were detected most often in the winter, and blue- and fin whale sounds were detected most often on the northern hydrophones. Sounds from seismic airguns were recorded frequently, particularly during summer, from locations over 3000 km from this array. Whales were detected by these hydrophones despite its location in a very remote part of the Atlantic Ocean that has traditionally been difficult to survey.     

Vocal characteristics of pygmy blue whales and their change over time

Vocal characteristics of pygmy blue whales of the eastern Indian Ocean population were analyzed using data from a hydroacoustic station deployed off Cape Leeuwin in Western Australia as part of the Comprehensive Nuclear-Test-Ban Treaty monitoring network, from two acoustic observatories of the Australian Integrated Marine Observing System, and from individual sea noise loggers deployed in the Perth Canyon. These data have been collected from 2002 to 2010, inclusively. It is shown that the themes of pygmy blue whale songs consist of ether three or two repeating tonal sounds with harmonics. The most intense sound of the tonal theme was estimated to correspond to a source level of 179 ± 2 dB re 1 μPa at 1 m measured for 120 calls from seven different animals. Short-duration calls of impulsive downswept sound from pygmy blue whales were weaker with the source level estimated to vary between 168 to 176 dB. A gradual decrease in the call frequency with a mean rate estimated to be 0.35 ± 0.3 Hz/year was observed over nine years in the frequency of the third harmonic of tonal sound 2 in the whale song theme, which corresponds to a negative trend of about 0.12 Hz/year in the call fundamental frequency.     

Steady inter and intra-annual decrease in the vocalization frequency of Antarctic blue whales

Time averaged narrow-band noise near 27 Hz produced by vocalizations of many distant Antarctic blue whales intensifies seasonally from early February to late October in the ocean off Australia's South West. Spectral characteristics of long term patterns in this noise band were analyzed using ambient noise data collected at the Comprehensive Nuclear-Test-Ban Treaty hydroacoustic station off Cape Leeuwin, Western Australia over 2002-2010. Within 7 day averaged noise spectra derived from 4096-point FFT (～0.06 Hz frequency resolution), the -3-dB width of the spectral peak from the upper tone of Antarctic blue whale vocalization was about 0.5 Hz. The spectral frequency peak of this tonal call was regularly but not gradually decreasing over the 9 years of observation from ～27.7 Hz in 2002 to ～26.6 Hz in 2010. The average frequency peak steadily decreased at a greater rate within a season at 0.4-0.5 Hz/season but then in the next year recovered to approximately the mean value of the previous season. A regression analysis showed that the interannual decrease rate of the peak frequency of the upper tonal call was 0.135 ± 0.003 Hz/year over 2002-2010 (R(2) ≈ 0.99). Possible causes of such a decline in the whale vocalization frequency are considered.     

Classification of behavior using vocalizations of Pacific white-sided dolphins (Lagenorhynchus obliquidens)

Surface behavior and concurrent underwater vocalizations were recorded for Pacific white-sided dolphins in the Southern California Bight (SCB) over multiple field seasons spanning 3 years. Clicks, click trains, and pulsed calls were counted and classified based on acoustic measurements, leading to the identification of 19 key call features used for analysis. Kruskal-Wallis tests indicated that call features differ significantly across behavioral categories. Previous work had discovered two distinctive click Types (A and B), which may correspond to known subpopulations of Pacific white-side dolphins in the Southern California Bight; this study revealed that animals producing these different click types also differ in both their behavior and vocalization patterns. Click Type A groups were predominantly observed slow traveling and milling, with little daytime foraging, while click Type B groups were observed traveling and foraging. These behavioral differences may be characteristic of niche partitioning by overlapping populations; coupled with differences in vocalization patterns, they may signify that these subpopulations are cryptic species. Finally, random forest decision trees were used to classify behavior based on vocalization data, with rates of correct classification up to 86%, demonstrating the potential for the use of vocalization patterns to predict behavior.     

Fundamental frequency and tracheal pressure during three types of vocalizations elicited from anesthetized dogs

Electrical stimulation of the midbrain was used to elicit a variety of vocalizations from six anesthetized dogs. This study was conducted to investigate the ranges of and relationships between fundamental frequency of the vocalizations (F₀) and tracheal pressure (P_t) produced during the vocalizations. The vocalizations were described according to type (growl, howl, and whine); F₀ and P_t, as well as patterns of laryngeal muscle activity, were examined for each vocalization type. Natural-sounding growl and howl vocalizations were elicited from five dogs; three dogs also produced whines. With few exceptions, F₀ was categorically different for the three vocalization types (low for growls, average for howls, very high for whines). P_t values overlapped for the three vocalization types, although, on average, howls were produced with greater P_t than growls. Patterns and degrees of laryngeal muscle activity varied across and within vocalization types, but general findings were consistent with the presumed function of most of the muscles. Laryngeal muscle activity may help explain some of the variability in the acoustic and aerodynamic data.