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.

Echolocating Daubenton's bats are resilient to broadband, ultrasonic masking noise during active target approaches.

Echolocating bats hunt prey on the wing under conditions of poor lighting by emission of loud calls and subsequent auditory processing of weak returning echoes. To do so, they need adequate echo-to-noise ratios (ENRs) to detect and distinguish target echoes from masking noise. Early obstacle avoidance experiments report high resilience to masking in free-flying bats, but whether this is due to spectral or spatiotemporal release from masking, advanced auditory signal detection or an increase in call amplitude (Lombard effect) remains unresolved. We hypothesized that bats with no spectral, spatial or temporal release from masking noise defend a certain ENR via a Lombard effect. We trained four bats (Myotis daubentonii) to approach and land on a target that broadcasted broadband noise at four different levels. An array of seven microphones enabled acoustic localization of the bats and source level estimation of their approach calls. Call duration and peak frequency did not change, but average call source levels (SLRMS, at 0.1 m as dB re. 20 µPa) increased, from 112 dB in the no-noise treatment, to 118 dB (maximum 129 dB) at the maximum noise level of 94 dB re. 20 µPa root mean square. The magnitude of the Lombard effect was small (0.13 dB SLRMS dB-1 of noise), resulting in mean broadband and narrowband ENRs of -11 and 8 dB, respectively, at the highest noise level. Despite these poor ENRs, the bats still performed echo-guided landings, making us conclude that they are very resilient to masking even when they cannot avoid it spectrally, spatially or temporally.

Oilbirds.

A Quick guide to oilbirds: nocturnal birds found only in Neotropical rainforests that, rather like many bat species, live in caves where they use echolocation for orientation.

Oilbirds produce echolocation signals beyond their best hearing range and adjust signal design to natural light conditions.

Oilbirds are active at night, foraging for fruits using keen olfaction and extremely light-sensitive eyes, and echolocate as they leave and return to their cavernous roosts. We recorded the echolocation behaviour of wild oilbirds using a multi-microphone array as they entered and exited their roosts under different natural light conditions. During echolocation, the birds produced click bursts (CBs) lasting less than 10 ms and consisting of a variable number (2-8) of clicks at 2-3 ms intervals. The CBs have a bandwidth of 7-23 kHz at -6 dB from signal peak frequency. We report on two unique characteristics of this avian echolocation system. First, oilbirds reduce both the energy and number of clicks in their CBs under conditions of clear, moonlit skies, compared with dark, moonless nights. Second, we document a frequency mismatch between the reported best frequency of oilbird hearing (approx. 2 kHz) and the bandwidth of their echolocation CBs. This unusual signal-to-sensory system mismatch probably reflects avian constraints on high-frequency hearing but may still allow oilbirds fine-scale, close-range detail resolution at the upper extreme (approx. 10 kHz) of their presumed hearing range. Alternatively, oilbirds, by an as-yet unknown mechanism, are able to hear frequencies higher than currently appreciated.

Fast sensory-motor reactions in echolocating bats to sudden changes during the final buzz and prey intercept.

Echolocation is an active sense enabling bats and toothed whales to orient in darkness through echo returns from their ultrasonic signals. Immediately before prey capture, both bats and whales emit a buzz with such high emission rates (≥ 180 Hz) and overall duration so short that its functional significance remains an enigma. To investigate sensory-motor control during the buzz of the insectivorous bat Myotis daubentonii, we removed prey, suspended in air or on water, before expected capture. The bats responded by shortening their echolocation buzz gradually; the earlier prey was removed down to approximately 100 ms (30 cm) before expected capture, after which the full buzz sequence was emitted both in air and over water. Bats trawling over water also performed the full capture behavior, but in-air capture motions were aborted, even at very late prey removals (<20 ms = 6 cm before expected contact). Thus, neither the buzz nor capture movements are stereotypical, but dynamically adapted based on sensory feedback. The results indicate that echolocation is controlled mainly by acoustic feedback, whereas capture movements are adjusted according to both acoustic and somatosensory feedback, suggesting separate (but coordinated) central motor control of the two behaviors based on multimodal input. Bat echolocation, especially the terminal buzz, provides a unique window to extremely fast decision processes in response to sensory feedback and modulation through attention in a naturally behaving animal.

Echolocation in Oilbirds and swiftlets.

The discovery of ultrasonic bat echolocation prompted a wide search for other animal biosonar systems, which yielded, among few others, two avian groups. One, the South American Oilbird (Steatornis caripensis: Caprimulgiformes), is nocturnal and eats fruit. The other is a selection of diurnal, insect-eating swiftlets (species in the genera Aerodramus and Collocalia: Apodidae) from across the Indo-Pacific. Bird echolocation is restricted to lower frequencies audible to humans, implying a system of poorer resolution than the ultrasonic (>20 kHz) biosonar of most bats and toothed whales. As such, bird echolocation has been labeled crude or rudimentary. Yet, echolocation is found in at least 16 extant bird species and has evolved several times in avian lineages. Birds use their syringes to produce broadband click-type biosonar signals that allow them to nest in dark caves and tunnels, probably with less predation pressure. There are ongoing discrepancies about several details of bird echolocation, from signal design to the question about whether echolocation is used during foraging. It remains to be seen if bird echolocation is as sophisticated as that of tongue-clicking rousette bats. Bird echolocation performance appears to be superior to that of blind humans using signals of notable similarity. However, no apparent specializations have been found so far in the birds' auditory system (from middle ear to higher processing centers). The advent of light-weight recording equipment and custom software for examining signals and reconstructing flight paths now provides the potential to study the echolocation behavior of birds in more detail and resolve such issues.

Intensity and directionality of bat echolocation signals.

The paper reviews current knowledge of intensity and directionality of bat echolocation signals. Recent studies have revealed that echolocating bats can be much louder than previously believed. Bats previously dubbed "whispering" can emit calls with source levels up to 110 dB SPL at 10 cm and the louder open space hunting bats have been recorded at above 135 dB SPL. This implies that maximum emitted intensities are generally 30 dB or more above initial estimates. Bats' dynamic control of acoustic features also includes the intensity and directionality of their sonar calls. Aerial hawking bats will increase signal directionality in the field along with intensity thus increasing sonar range. During the last phase of prey pursuit, vespertilionid bats broaden their echolocation beam considerably, probably to counter evasive maneuvers of eared prey. We highlight how multiple call parameters (frequency, duration, intensity, and directionality of echolocation signals) in unison define the search volume probed by bats and in turn how bats perceive their surroundings. Small changes to individual parameters can, in combination, drastically change the bat's perception, facilitating successful navigation and food acquisition across a vast range of ecological niches. To better understand the function of echolocation in the natural habitat it is critical to determine multiple acoustic features of the echolocation calls. The combined (interactive) effects, not only of frequency and time parameters, but also of intensity and directionality, define the bat's view of its acoustic scene.

New model for gain control of signal intensity to object distance in echolocating bats.

Echolocating bats emit ultrasonic calls and listen for the returning echoes to orient and localize prey in darkness. The emitted source level, SL (estimated signal intensity 10 cm from the mouth), is adjusted dynamically from call to call in response to sensory feedback as bats approach objects. A logarithmic relationship of SL=20 log(10)(x), i.e. 6 dB output reduction per halving of distance, x, has been proposed as a model for the relationship between emitted intensity and object distance, not only for bats but also for echolocating toothed whales. This logarithmic model suggests that the approaching echolocator maintains a constant intensity impinging upon the object, but it also implies ever-increasing source levels with distance, a physical and biological impossibility. We developed a new model for intensity compensation with an exponential rise to the maximum source level: SL=SL(max)-ae(-)(bx). In addition to providing a method for estimating maximum output, the new model also offers a tool for estimating a minimum detection distance where intensity compensation starts. We tested the new exponential model against the 'conventional' logarithmic model on data from five bat species. The new model performed better in 77% of the trials and as good as the conventional model in the rest (23%). We found much steeper rates of compensation when fitting the model to individual rather than pooled data, with slopes often steeper than -20 dB per halving of distance. This emphasizes the importance of analyzing individual events. The results are discussed in light of habitat constraints and the interaction between bats and their eared prey.

Echolocation call intensity and directionality in flying short-tailed fruit bats, Carollia perspicillata (Phyllostomidae).

The directionality of bat echolocation calls defines the width of bats' sonar "view," while call intensity directly influences detection range since adequate sound energy must impinge upon objects to return audible echoes. Both are thus crucial parameters for understanding biosonar signal design. Phyllostomid bats have been classified as low intensity or "whispering bats," but recent data indicate that this designation may be inaccurate. Echolocation beam directionality in phyllostomids has only been measured through electrode brain-stimulation of restrained bats, presumably excluding active beam control via the noseleaf. Here, a 12-microphone array was used to measure echolocation call intensity and beam directionality in the frugivorous phyllostomid, Carollia perspicillata, echolocating in flight. The results showed a considerably narrower beam shape (half-amplitude beam angles of approximately 16° horizontally and 14° vertically) and louder echolocation calls [source levels averaging 99 dB sound pressure level (SPL) root mean square] for C. perspicillata than was found for this species when stationary. This suggests that naturally behaving phyllostomids shape their sound beam to achieve a longer and narrower sonar range than previously thought. C. perspicillata orient and forage in the forest interior and the narrow beam might be adaptive in clutter, by reducing the number and intensity of off-axis echoes.

Intense echolocation calls from two 'whispering' bats, Artibeus jamaicensis and Macrophyllum macrophyllum (Phyllostomidae).

Bats use echolocation to exploit a variety of habitats and food types. Much research has documented how frequency-time features of echolocation calls are adapted to acoustic constraints imposed by habitat and prey but emitted sound intensities have received little attention. Bats from the family of Phyllostomidae have been categorised as low intensity (whispering) gleaners, assumed to emit echolocation calls with low source levels (approximately 70 dB SPL measured 10 cm from the bat's mouth). We used a multi-microphone array to determine intensities emitted from two phyllostomid bats from Panamá with entirely different foraging strategies. Macrophyllum macrophyllum hunts insects on the wing and gaffs them with its tail membrane and feet from or above water surfaces whereas Artibeus jamaicensis picks fruit from vegetation with its mouth. Recordings were made from bats foraging on the wing in a flight room. Both species emitted surprisingly intense signals with maximum source levels of 105 dB SPL r.m.s. for M. macrophyllum and 110 dB SPL r.m.s. for A. jamaicensis, hence much louder than a ;whisper'. M. macrophyllum was consistently loud (mean source level 101 dB SPL) whereas A. jamaicensis showed a much more variable output, including many faint calls and a mean source level of 96 dB SPL. Our results support increasing evidence that echolocating bats in general are much louder than previously thought. We discuss the importance of loud calls and large output flexibility for both species in an ecological context.