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.

Albert Feng: father, friend, scientist, innovator (1944-2021).

Albert S.-H. Feng was an outstanding family man and brilliant scientist, with a creative mind, a gift for dealing with people of all types, and a warm, personable demeanor. He was blessed with many talents, making him a sought-after colleague and collaborator. His love for people and travel took him to many destinations around the world where he studied the neuroethology of frog and bat communication both in the field and in the lab. He has made many fundamental contributions to our understanding of the vertebrate auditory system. These include characterizing the "delay-tuned" neurons in the bat midbrain underlying target detection, and in discovering several terrestrial amphibians in which the upper limit of hearing extends well into the ultrasonic range, forever changing our long-held perception of frogs as "low-frequency animals".

Acoustic camouflage increases with body size and changes with bat echolocation frequency range in a community of nocturnally active Lepidoptera.

Body size is an important trait in predator-prey dynamics as it is often linked to detection, as well as the success of capture or escape. Larger prey, for example, often runs higher risk of detection by their predators, which imposes stronger selection on their anti-predator traits compared to smaller prey. Nocturnal Lepidoptera (moths) vary strongly in body size, which has consequences for their predation risk, as bigger moths return stronger echoes for echolocating bats. To compensate for increased predation risk, larger moths are therefore expected to have improved anti-predator defences. Moths are covered by different types of scales, which for a few species are known to absorb ultrasound, thus providing acoustic camouflage. Here, we assessed whether moths differ in their acoustic camouflage in a size-dependent way by focusing on their body scales and the different frequency ranges used by bats. We used a sonar head to measure 3D echo scans of a total of 111 moth specimens across 58 species, from eight different families of Lepidoptera. We scanned all the specimens and related their echo-acoustic target strength to various body size measurements. Next, we removed the scales covering the thorax and abdomen and scanned a subset of specimens again to assess the sound absorptive properties of these scales. Comparing intact specimens with descaled specimens, we found almost all species to absorb ultrasound, reducing detection risk on average by 8%. Furthermore, the sound absorptive capacities of body scales increased with body size suggesting that larger species benefit more from acoustic camouflage. The size-dependent effect of camouflage was in particular pronounced for the higher frequencies (above 29 kHz), with moth species belonging to large-bodied families consequently demonstrating similar target strengths compared to species from small-bodied families. Finally, we found the families to differ in frequency range that provided the largest reduction in detection risk, which may be related to differences in predation pressure and predator communities of these families. In general, our findings have important implications for predator-prey interactions across eco-evolutionary timescales and may suggest that acoustic camouflage played a role in body size evolution of nocturnally active Lepidoptera.

TrackUSF, a novel tool for automated ultrasonic vocalization analysis, reveals modified calls in a rat model of autism.

BACKGROUND: Various mammalian species emit ultrasonic vocalizations (USVs), which reflect their emotional state and mediate social interactions. USVs are usually analyzed by manual or semi-automated methodologies that categorize discrete USVs according to their structure in the frequency-time domains. This laborious analysis hinders the effective use of USVs as a readout for high-throughput analysis of behavioral changes in animals. RESULTS: Here we present a novel automated open-source tool that utilizes a different approach towards USV analysis, termed TrackUSF. To validate TrackUSF, we analyzed calls from different animal species, namely mice, rats, and bats, recorded in various settings and compared the results with a manual analysis by a trained observer. We found that TrackUSF detected the majority of USVs, with less than 1% of false-positive detections. We then employed TrackUSF to analyze social vocalizations in Shank3-deficient rats, a rat model of autism, and revealed that these vocalizations exhibit a spectrum of deviations from appetitive calls towards aversive calls. CONCLUSIONS: TrackUSF is a simple and easy-to-use system that may be used for a high-throughput comparison of ultrasonic vocalizations between groups of animals of any kind in any setting, with no prior assumptions.

Ultrasound avoidance by flying antlions (Myrmeleontidae).

The acoustic arms race between insectivorous bats and their invertebrate prey has led to the convergent evolution of ultrasound hearing in seven orders of nocturnal insects. Upon hearing the echolocation calls of an approaching bat, such insects take defensive action. Here, we document a previously unknown sense of ultrasound hearing and phonotactic flight behaviour in the neuropteran family Myrmeleontidae (antlions). The antlion Myrmeleon hyalinus was presented with sound pulses at ultrasonic frequencies used by echolocating bats and its response thresholds in tethered flight determined. Behaviours included abdominal twitches, wing flicks, brief pauses in flight and flight cessation. Such behaviours create erratic evasive flight manoeuvres in other eared insects, particularly mantids and lacewings. Antlions responded best to ultrasound between 60 and 80 kHz (75 dB peSPL at 80 kHz), showing response thresholds similar to those of the related lacewings (Neuroptera, Chrysopidae). Yet, at lower ultrasonic frequencies (20-50 kHz), antlions were far less sensitive than lacewings. Based on calculated response distances, we conclude that antlions respond only after having been detected by bats rather than using early evasive flights. We argue that the high response threshold for low-frequency ultrasound is adaptive for an insect that is mainly active close to and within vegetation, because a behavioural response to the lower ultrasonic frequencies used by high-flying bats would result in evasive action in the absence of actual predation risk.

Modeling bat prey capture in echolocating bats: The feasibility of reactive pursuit.

Echolocating bats are the only mammals engaging in airborne pursuit. In this paper, we implement a reactive model of sonar based prey pursuit in bats. Our simulations include a realistic prey localization mechanism as well as a model of the bat's motor behavior. In contrast to previous work, our model incorporates bats' ability to execute rapid saccadic scanning motions keeping the prey within its field of view. Decoupling the flight direction from the gaze direction allows our model to capture erratically moving prey using reactive control. We conclude that the rapid shifts in gaze direction allow bats to deal with the narrow field of view provided by their sonar system.

Target shape perception and clutter rejection use the same mechanism in bat sonar.

Big brown bats (Eptesicus fuscus) emit frequency-modulated (FM) biosonar sounds containing two or more harmonic sweeps. Echoes from frontally located targets arrive with first and second harmonics intact, leading to focused delay images. Echoes from offside or distant objects arrive with the second harmonic relatively weaker (lowpass-filtered), leading to defocused images, which prevents their clutter interference effects (Bates et al. J Exp Biol 214:394-401, 2011). Realistic targets contain several glints at slightly different distances and reflect several echoes at correspondingly different delays. The bat registers the delay of the nearest glint's echoes in the time domain. The delays of echoes from the farther glints are registered in the frequency domain, from interference nulls in the spectrum. Lowpass-filtering of echoes directly affects the image of the nearest glint by defocusing the delay image. However, lowpass-filtering also is superimposed on the interference spectrum used to register the farther glints, which distorts the pattern of interference nulls, defocusing the farther glints inversely, in the spectral domain, before they are perceived as delays. Differences in blurring between time-domain and frequency-domain parts of images identifies separate computational paths to perceptually reconstruct objects and prevent interference from off-side or distant clutter.

Big brown bats (Eptesicus fuscus) emit intense search calls and fly in stereotyped flight paths as they forage in the wild.

The big brown bat, Eptesicus fuscus, uses echolocation for orientation and foraging, and scans its surroundings by aiming its sonar beam at obstacles and prey. All call parameters are highly adaptable and determine the bat's acoustic field of view and hence its perception of the echo scene. The intensity (source level) and directionality of the emitted calls directly contribute to the bat's acoustic field of view; however, the source level and directionality of the big brown bat's sonar signals have not been measured in the field. In addition, for bats, navigation and prey capture require that they process several streams of acoustic information. By using stereotypic flight paths in known areas, bats may be able to reduce the sensory processing load for orientation and therefore allocate echo processing resources to prey. Here we recorded the echolocation calls from foraging E. fuscus in the field with a microphone array and estimated call intensity and directionality, based on reconstructed flight trajectories. The source levels were intense with an average maximum source level of 138 dB (root mean square re. 20 µPa at 0.1 m). Furthermore, measurements taken from a subset of calls indicate that the echolocation signals in the field may be more directional than estimated in the laboratory (half-amplitude angle 30 deg at 35 kHz). We also observed that E. fuscus appear to follow stereotypic flight paths, and propose that this could be a strategy to optimize foraging efficiency by minimizing the sensory processing load.

Ambient noise causes independent changes in distinct spectro-temporal features of echolocation calls in horseshoe bats.

One of the most efficient mechanisms to optimize signal-to-noise ratios is the Lombard effect - an involuntary rise in call amplitude due to ambient noise. It is often accompanied by changes in the spectro-temporal composition of calls. We examined the effects of broadband-filtered noise on the spectro-temporal composition of horseshoe bat echolocation calls, which consist of a constant-frequency component and initial and terminal frequency-modulated components. We found that the frequency-modulated components became larger for almost all noise conditions, whereas the bandwidth of the constant-frequency component increased only when broadband-filtered noise was centered on or above the calls' dominant or fundamental frequency. This indicates that ambient noise independently modifies the associated acoustic parameters of the Lombard effect, such as spectro-temporal features, and could significantly affect the bat's ability to detect and locate targets. Our findings may be of significance in evaluating the impact of environmental noise on echolocation behavior in bats.

Bats integrate multiple echolocation and flight tactics to track prey.

The ability of "target tracking," such as keeping a target object in sight, is crucial for various activities. However, most sensing systems experience a certain degree of delay due to information processing, which challenges accurate target tracking. The long history of studies on animal behavior has revealed several tactics for it, although a systematic understanding of how individual tactics are combined into a strategy has not been reached. This study demonstrates a multifaceted tracking strategy in animals, which mitigates the adverse delay effects with small implementation costs. Using an active-sensing bat to measure their sensing state while chasing natural prey, we found that bats use a tracking strategy by combining multiple echolocation and flight tactics. The three echolocation tactics, namely the predictive control of sensing direction accompanied by adjusting the sensing rate and angular range, produce a direct compensation effect. Simultaneously, the flight tactic, the counter maneuver, assists echolocation by stabilizing the target direction. Our simulation results demonstrate that these combined tactics improve tracking accuracy over a wide range of delay constraints. In addition, a concise rule based on the angular velocity between bats and targets explains how bats control these tactics, suggesting that bats successfully reduce the burden of multitasking management. Our findings reveal the sophisticated strategy in animals' tracking systems and provide insights into understanding and developing efficiently integrated strategies in target tracking across various disciplines.

Habitat suitability mapping using logistic regression analysis of long-term bioacoustic bat survey dataset in the Cassadaga Creek watershed (USA).

Bat species show global ecological importance, yet their numbers are declining worldwide. Understanding bat-habitat interactions is crucial in terms of developing effective conservation plans. In an effort to model bat habitat suitability in the Cassadaga Creek watershed, long-term bioacoustic bat data (spanning 2009-2020) was compiled, georeferenced and statistically analyzed using logistic regression techniques. In total, 1600 bat occurrence records from five species of bat (559 Eptesicus fuscus, 560 Lasionycteris noctivagans, 143 Lasiurus borealis, 260 Lasiurus cinereus, and 78 Myotis lucifugus) were paired with pseudo-absence points to study the relationship between bat calling behavior and land cover. All bats but Myotis lucifugus had a statistically significant relationship with forested land cover, and all bats had negative interactions with agricultural habitats. Geospatial data was coupled with the statistical output to create maps of habitat suitability and echolocation calling density. This work provides a model that can be employed worldwide to evaluate bat habitat needs or patterns in echolocation behavior. Future research will incorporate a more recently collected dataset that is of greater geographic diversity with a larger number of environmental variables in the species distribution model.

An Efficient Neural Network Design Incorporating Autoencoders for the Classification of Bat Echolocation Sounds.

Bats are widely distributed around the world, have adapted to many different environments and are highly sensitive to changes in their habitat, which makes them essential bioindicators of environmental changes. Passive acoustic monitoring over long durations, like months or years, accumulates large amounts of data, turning the manual identification process into a time-consuming task for human experts. Automated acoustic monitoring of bat activity is therefore an effective and necessary approach for bat conservation, especially in wind energy applications, where flying animals like bats and birds have high fatality rates. In this work, we provide a neural-network-based approach for bat echolocation pulse detection with subsequent genus classification and species classification under real-world conditions, including various types of noise. Our supervised model is supported by an unsupervised learning pipeline that uses autoencoders to compress linear spectrograms into latent feature vectors that are fed into a UMAP clustering algorithm. This pipeline offers additional insights into the data properties, aiding in model interpretation. We compare data collected from two locations over two consecutive years sampled at four heights (10 m, 35 m, 65 m and 95 m). With sufficient data for each labeled bat class, our model is able to comprehend the full echolocation soundscape of a species or genus while still being computationally efficient and simple by design. Measured classification F1 scores in a previously unknown test set range from 92.3% to 99.7% for species and from 94.6% to 99.4% for genera.

The Curved Openspace Algorithm and a Spike-Latency Model for Sonar-Based Obstacle Avoidance.

The rapid control of a sonar-guided vehicle to pursue a goal while avoiding obstacles has been a persistent research topic for decades. Taking into account the limited field-of-view of practical sonar systems and vehicle kinematics, we propose a neural model for obstacle avoidance that maps the 2-D sensory space into a 1-D motor space and evaluates motor actions while combining obstacles and goal information. A two-stage winner-take-all (WTA) mechanism is used to select the final steering action. To avoid excessive scanning of the environment, an attentional system is proposed to control the directions of sonar pings for efficient, task-driven, sensory data collection. A mobile robot was used to test the proposed model navigating through a cluttered environment using a narrow field-of-view sonar system. We further propose a spiking neural model using spike-timing representations, a spike-latency memory, and a "race-to-first-spike" WTA circuit.

Unusual echolocation behaviour of the common sword-nosed bat Lonchorhina aurita: an adaptation to aerial insectivory in a phyllostomid bat?

Most insectivorous bat species in the Neotropical family Phyllostomidae glean insects from ground, water or vegetation surfaces. They use similar and stereotypical echolocation calls that are generally very short (less than 1-3 ms), multi-harmonic and frequency-modulated (FM). By contrast, the common sword-nosed bat, Lonchorhina aurita, which has the longest noseleaf in the entire phyllostomid family, produces distinctly different echolocation calls. They are composed of a constant frequency (CF) component with a peak frequency of 45 kHz, followed by a short FM down-sweep at the end. With a mean call duration of 6.6 ms (max. 8.7 ms) when flying in the open they have the longest echolocation calls reported from phyllostomid bats. In cluttered environments, the CF-component is very short. In open habitats, however, L. aurita can emit pure CF-calls without the terminal FM-component. We also recorded in the field a distinct terminal group that closely resembles the feeding buzzes of aerial hawking bat species from other bat families. This is the first time the echolocation call design of L. aurita is presented. In addition, we contrast the echolocation behaviour of individuals flying in open and confined situations. Our results suggest that the unique echolocation system of L. aurita represents an adaptation to aerial hawking, a very unusual hunting mode within the phyllostomid family.

Behavioral flexibility of the trawling long-legged bat, Macrophyllum macrophyllum (Phyllostomidae).

We assessed the behavioral flexibility of the trawling long-legged bat, Macrophyllum macrophyllum (Phyllostomidae) in flight cage experiments by exposing it to prey suspended from nylon threads in the air and to food placed onto the water surface at varying distances to clutter-producing background (water plants). The bat revealed flexibility in foraging mode and caught prey in the air (aerial hawking) and from the water surface (trawling). M. macrophyllum was constrained in finding food very near to and within clutter. As echolocation was the prime sensory mode used by M. macrophyllum for detection and localization of food, the bat might have been unable to perceive sufficient information from prey near clutter as background echoes from the water plant increasingly overlapped with echoes from food. The importance of echolocation for foraging is reflected in a stereotypic call pattern of M. macrophyllum that resembles other aerial insectivorous and trawling bats with a pronounced terminal phase (buzz) prior to capture attempts. Our findings contrast studies of other phyllostomid bats that glean prey very near or from vegetation, often using additional sensory cues, such as prey-produced noise, to find food and that lack a terminal phase in echolocation behavior. In M. macrophyllum, acoustic characteristics of its foraging habitat have shaped its sonar system more than phylogeny.