Elephant Vocalizations Research Articles

Information content and acoustic structure of male African elephant social rumbles.

Until recently, the prevailing theory about male African elephants (Loxodonta africana) was that, once adult and sexually mature, males are solitary and targeted only at finding estrous females. While this is true during the state of ‘musth’ (a condition characterized by aggressive behavior and elevated androgen levels), ‘non-musth’ males exhibit a social system seemingly based on companionship, dominance and established hierarchies. Research on elephant vocal communication has so far focused on females, and very little is known about the acoustic structure and the information content of male vocalizations. Using the source and filter theory approach, we analyzed social rumbles of 10 male African elephants. Our results reveal that male rumbles encode information about individuality and maturity (age and size), with formant frequencies and absolute fundamental frequency values having the most informative power. This first comprehensive study on male elephant vocalizations gives important indications on their potential functional relevance for male-male and male-female communication. Our results suggest that, similar to the highly social females, future research on male elephant vocal behavior will reveal a complex communication system in which social knowledge, companionship, hierarchy, reproductive competition and the need to communicate over long distances play key roles.

Towards an automated acoustic detection system for free-ranging elephants

The human–elephant conflict is one of the most serious conservation problems in Asia and Africa today. The involuntary confrontation of humans and elephants claims the lives of many animals and humans every year. A promising approach to alleviate this conflict is the development of an acoustic early warning system. Such a system requires the robust automated detection of elephant vocalizations under unconstrained field conditions. Today, no system exists that fulfils these requirements. In this paper, we present a method for the automated detection of elephant vocalizations that is robust to the diverse noise sources present in the field. We evaluate the method on a data-set recorded under natural field conditions to simulate a real-world scenario. The proposed method outperformed existing approaches and robustly and accurately detected elephants. It thus can form the basis for a future automated early warning system for elephants. Furthermore, the method may be a useful tool for scientists in bioacoustics for the study of wildlife recordings.

Vocal stress associated with a translocation of a family herd of African elephants (Loxodonta africana) in the Kruger National Park, South Africa

We used vocal indicators to examine the effect of a translocation of an African elephant family herd within the Kruger National Park. These animals were moved 300 km from their home range, but returned unaided to this range within 23 days. We found that translocation resulted in a change in the mean fundamental frequency of low-frequency elephant vocalizations, known as rumbles. The rumbles increased significantly in pitch compared with pre-translocation levels during the 23 days the animals spent outside their normal home range. Mean fundamental frequency returned close to pre-translocation level by the time the animals had navigated their way back to their previous home range. Raised pitch is known to be an indicator of stress in humans and other animals. The observed acoustic results are consistent with a physiological measure of stress, faecal glucocorticoid metabolite levels, which were monitored from the same animals during the study and have already been reported elsewhere. To our knowledge, this is the first report of prolonged monitoring of vocal stress response in free-ranging animals. Measuring behavioural responses, such as vocalizations, may provide an objective non-invasive method for assessing stress. This could help in determining the effects that particular management actions might have on elephants.

Visualizing Sound Emission of Elephant Vocalizations: Evidence for Two Rumble Production Types

Recent comparative data reveal that formant frequencies are cues to body size in animals, due to a close relationship between formant frequency spacing, vocal tract length and overall body size. Accordingly, intriguing morphological adaptations to elongate the vocal tract in order to lower formants occur in several species, with the size exaggeration hypothesis being proposed to justify most of these observations. While the elephant trunk is strongly implicated to account for the low formants of elephant rumbles, it is unknown whether elephants emit these vocalizations exclusively through the trunk, or whether the mouth is also involved in rumble production. In this study we used a sound visualization method (an acoustic camera) to record rumbles of five captive African elephants during spatial separation and subsequent bonding situations. Our results showed that the female elephants in our analysis produced two distinct types of rumble vocalizations based on vocal path differences: a nasally- and an orally-emitted rumble. Interestingly, nasal rumbles predominated during contact calling, whereas oral rumbles were mainly produced in bonding situations. In addition, nasal and oral rumbles varied considerably in their acoustic structure. In particular, the values of the first two formants reflected the estimated lengths of the vocal paths, corresponding to a vocal tract length of around 2 meters for nasal, and around 0.7 meters for oral rumbles. These results suggest that African elephants may be switching vocal paths to actively vary vocal tract length (with considerable variation in formants) according to context, and call for further research investigating the function of formant modulation in elephant vocalizations. Furthermore, by confirming the use of the elephant trunk in long distance rumble production, our findings provide an explanation for the extremely low formants in these calls, and may also indicate that formant lowering functions to increase call propagation distances in this species'.

Science Podcast

The show includes the quantum mechanics of black holes, analyzing elephant vocalizations, and chasing down the last of the polio virus.

Automatic classification and speaker identification of African elephant (Loxodonta africana) vocalizations

A hidden Markov model (HMM) system is presented for automatically classifying African elephant vocalizations. The development of the system is motivated by successful models from human speech analysis and recognition. Classification features include frequency-shifted Mel-frequency cepstral coefficients (MFCCs) and log energy, spectrally motivated features which are commonly used in human speech processing. Experiments, including vocalization type classification and speaker identification, are performed on vocalizations collected from captive elephants in a naturalistic environment. The system classified vocalizations with accuracies of 94.3% and 82.5% for type classification and speaker identification classification experiments, respectively. Classification accuracy, statistical significance tests on the model parameters, and qualitative analysis support the effectiveness and robustness of this approach for vocalization analysis in nonhuman species.

Comparison and evaluation of animal vocalization enhancement techniques

Animal vocalization recordings are often corrupted by wideband background noise and interfering signals; however, signal processing methods for enhancement of these waveforms has not received as much attention in the literature as has human speech enhancement. In order to improve vocalization intelligibility and quality, a variety of enhancement methods taken from the field of speech processing are investigated here. These techniques range from traditional approaches such as spectral subtraction to more advanced ones such as Ephraim Malah log-spectral estimation and wavelet denoising. Results, measured by improvement in signal-to-noise ratio (SNR) and subjective perceptual tests, are given for several noise environments. Signal enhancement is demonstrated across a variety of species, including African elephant, Beluga whale, and ortolan bunting vocalizations. [Work supported by NSF.]

Seismic waves from elephant vocalizations: A possible communication mode?

We conducted experiments with trained African elephants that show that low‐frequency elephant vocalizations produce Rayleigh waves. We model a potential range for these seismic waves, under ideal conditions, of c. 2 km. In appropriate conditions, surface waves from an elephant's infrasonic vocalizations might propagate further than airborne sound and provide advantages over acoustic communication. However, if we use the detection capabilities of the human ear as a benchmark for the signal‐detection thresholds of elephants, our estimates of attenuation and ambient seismic noise suggest that the seismic detection range is unlikely to exceed the acoustic detection range under normal atmospheric conditions. We conclude that elephants may benefit from seismic detection in circumstances where the range of acoustic communication is limited, or in cases where multimodal communication is advantageous. Given our current understanding, elephants are unlikely to rely on seismic waves as their primary mode for long‐range communication.

Elephant low-frequency vocalizations propagate in the ground and seismic playbacks of these vocalizations are detectable by wild African elephants (Loxodonta africana)

Seismic correlates of low-frequency vocalizations in African and Asian elephants propagate in the ground at different velocities, with the potential of traveling farther than their airborne counterparts. A semblance technique applied to linear moveouts on narrow-bandpass-filtered data, coupled with forward modeling, demonstrates that the complex waves observed are the interference of an air wave and a Rayleigh wave traveling at the appropriate velocities. The Rayleigh wave appears to be generated at or close to the elephant, either by coupling through the elephant’s body or through the air near the body to the ground. Low-frequency elephant vocalizations were reproduced seismically and played back to both a captive elephant and to elephant breeding herds in the wild, monitoring the elephants’ behavioral responses, spacing between herd members and time spent at the water hole as an index of heightened vigilance. Breeding herds detected and responded appropriately to seismically transmitted elephant warning calls. The captive studies promise to elucidate a vibrotactile threshold of sensitivity for the elephant foot. Elephants may benefit from the exploitation of seismic cues as an additional communication modality, thus expanding their signaling repertoire and extending their range of potential communication and eavesdropping beyond that possible with airborne sound.

Generalized perceptual features for animal vocalization classification

Two sets of generalized, perceptual-based features are investigated for use in classifying animal vocalizations. Since many species, especially mammals, share similar physical sound perception mechanisms which vary in size, two features sets commonly used in human speech processing, mel-frequency cepstral coefficients (MFCCs) and perceptual linear prediction (PLP) analysis, are modified for use in other species. One modification made to the feature extraction process is incorporating the frequency range of hearing and length of the basilar membrane of the animal in order to correctly determine the width and location of the critical band filters used for signal processing. Experimentally determined critical bands (equivalent rectangular bandwidth) and equal loudness curves (audiograms) can also be incorporated directly into the feature extraction process. Experiments are performed on African elephant (Loxodonta africana) vocalizations using a hidden Markov model (HMM) based classifier showing increased classification accuracy when using features sets based on the specific animals perceptual abilities compared to the original human perception-based feature sets.

Automatic type classification and speaker identification of african elephant (Loxodonta africana) vocalizations

This paper presents a system for automatically classifying African elephant vocalizations based on systems used for human speech recognition and speaker identification. The experiments are performed on vocalizations collected from captive elephants in a naturalistic environment. Features used for classification include Mel-Frequency Cepstral Coefficients (MFCCs) and log energy which are the most common features used in human speech processing. Since African elephants use lower frequencies than humans in their vocalizations, the MFCCs are computed using a shifted Mel-Frequency filter bank to emphasize the infrasound range of the frequency spectrum. In addition to these features, the use of less traditional features such as those based on fundamental frequency and the phase of the frequency spectrum is also considered. A Hidden Markov Model with Gaussian mixture state probabilities is used to model each type of vocalization. Vocalizations are classified based on type, speaker and estrous cycle. Experiments on continuous call type recognition, which can classify multiple vocalizations in the same utterance, are also performed. The long-term goal of this research is to develop a universal analysis framework and robust feature set for animal vocalizations that can be applied to many species.

Seismic properties of Asian elephant (Elephas maximus) vocalizations and locomotion.

Seismic and acoustic data were recorded simultaneously from Asian elephants (Elephas maximus) during periods of vocalizations and locomotion. Acoustic and seismic signals from rumbles were highly correlated at near and far distances and were in phase near the elephant and were out of phase at an increased distance from the elephant. Data analyses indicated that elephant generated signals associated with rumbles and "foot stomps" propagated at different velocities in the two media, the acoustic signals traveling at 309 m/s and the seismic signals at 248-264 m/s. Both types of signals had predominant frequencies in the range of 20 Hz. Seismic signal amplitudes considerably above background noise were recorded at 40 m from the generating elephants for both the rumble and the stomp. Seismic propagation models suggest that seismic waveforms from vocalizations are potentially detectable by instruments at distances of up to 16 km, and up to 32 km for locomotion generated signals. Thus, if detectable by elephants, these seismic signals could be useful for long distance communication.

Portable instrumentation for the recording, analysis, and playback of infrasonic animal vocalizations

Several bioacoustic experiments required an affordable system capable of field recording, real-time analysis, signal editing, and real-time playback of infrasonic animal vocalizations. The synthesis of several types of recording devices and signal processing software has resulted in the capacity to record infrasound and perform analysis even during rain, heat, and high humidity. Small portable microphones and recorders can record in the field from 3 Hz to 22 kHz. On-site analysis including real-time FFT, color spectrographic function, filtering, cross-correlation, and other functions can be facilitated with the use of any portable computer with 92 Mbytes RAM. Signal editing, including frequency, amplitude, and cut/paste can be accomplished with readily available music signal processing software. Real-time field playback from 10 Hz to 45 kHz can be acheived by using portable car audio speakers. This system was used for the real-time analysis of elephant vocalizations, which were edited and played back immediately during recording sessions. It has also been used in field research involving tiger, binturong, rhinoceros, and giraffe. Tigers and elephants appear to respond behaviorally to real-time playback of original and edited vocalizations. [Work supported by National Instruments, Momentum Data Systems, and Sonic Foundry.]

Signals and assessment in African elephants: evidence from playback experiments

A series of playback experiments using two elephant vocalizations, the ‘musth rumble’ and the ‘oestrous call’, was carried out in Amboseli National Park to examine signalling and assessment in African elephants, Loxodonta africana . In response to the musth rumble of a high-ranking male other musth males approached the speaker aggressively, whereas nonmusth males walked away from the stimulus. The call of an oestrous female, too, attracted musth males who approached the speaker rapidly, while nonmusth males listened and then walked away. Females listened and often showed considerable interest in the musth rumbles of males, approaching the speaker and sometimes responding by vocalizing and or secreting from the temporal glands. The experiments bear out earlier observational data and game theory predictions which suggest that by being in or out of musth a male may be conveying information about the relative value he places on contesting his dominance rank and his access to oestrous females. When not visibly in musth, a male may be indicating his intention not to contest access to oestrous females.

Long range seismic characteristics of Asian elephant (Elephus maximus) vocalizations and locomotion

Seismic components of Asian elephant (Elephus maximus) vocalizations and locomotion were recorded at a range of distances. To study the properties of sound attenuation in the ground, a string of four geophones was placed at approximately 30 m intervals from the elephants, up to 120 m from the signal source. Precise distances between sensors was measured with a differential GPS. Some trials were also conducted using microphone/geophone pairs at 30- and 120-m distances to compare attenuation in the air in relation to attenuation in the ground. Recordings were made with a four channel data acquisition system, processing signals simultaneously from the geophone string and the microphone/geophone pairs at the near and remote locations. Previous studies indicated that elephant vocalizations and locomotion produce Rayleigh waves, a type of ground surface wave. In this study, the long-range seismic characteristics of the signal sources is described. The propagation of elephant vocalizations in the air and ground is quantified at a range of distances up to 120 m and the rates of signal attenuation within each of the two media are compared.

Seismic transmission of elephant vocalizations and movement

Seismic signals were recorded during periods of vocalizations and movement by African and Asian elephants. Initially, vocalizations were detected in both microphone and geophone recordings. To test whether these sounds are propagated separately in the air and the ground, recordings were made with a four-channel data acquisition system, processing signals simultaneously from microphone/geophone pairs at near and remote points ranging from 20–130 m distances from the elephants. These recordings indicate that elephant vocalizations and movement produce Rayleigh waves, a type of ground surface wave. This paper quantifies the propagation of elephant vocalizations in the air and ground and compares the rates of signal attenuation within each of the two media. A normalized cross-correlation coefficient was performed on rumble and movement signals in the ground, resulting in a time lag between near and remote sensors corresponding to Rayleigh wave velocity. Rayleigh waves travel slower than airborne sound waves (approximately 250 m per second versus 340 m per second), but they can travel further and more reliably than those transmitted through the air. The presence of seismic properties in elephant vocalizations and movement suggests a possibility for their use in long distance communication.