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.

Molecular adaptations underlying high-frequency hearing in the brain of CF bats species.

BACKGROUND: The majority of bat species have developed remarkable echolocation ability, especially for the laryngeally echolocating bats along with high-frequency hearing. Adaptive evolution has been widely detected for the cochleae in the laryngeally echolocating bats, however, limited understanding for the brain which is the central to echolocation signal processing in the auditory perception system, the laryngeally echolocating bats brain may also undergo adaptive changes. RESULT: In order to uncover the molecular adaptations related with high-frequency hearing in the brain of laryngeally echolocating bats, the genes expressed in the brain of Rhinolophus ferrumequinum (CF bat) and Myotis pilosus (FM bat) were both detected and also compared. A total of 346,891 genes were detected and the signal transduction mechanisms were annotated by the most abundant genes, followed by the transcription. In hence, there were 3,088 DEGs were found between the two bat brains, with 1,426 highly expressed in the brain of R. ferrumequinum, which were significantly enriched in the neuron and neurodevelopmental processes. Moreover, we found a key candidate hearing gene, ADCY1, playing an important role in the R. ferrumequinum brain and undergoing adaptive evolution in CF bats. CONCLUSIONS: Our study provides a new insight to the molecular bases of high-frequency hearing in two laryngeally echolocating bats brain and revealed different nervous system activities during auditory perception in the brain of CF bats.

Characterising underwater noise and changes in harbour porpoise behaviour during the decommissioning of an oil and gas platform.

Many man-made marine structures (MMS) will have to be decommissioned in the coming decades. While studies on the impacts of construction of MMS on marine mammals exist, no research has been done on the effects of their decommissioning. The complete removal of an oil and gas platform in Scotland in 2021 provided an opportunity to investigate the response of harbour porpoises to decommissioning. Arrays of broadband noise recorders and echolocation detectors were used to describe noise characteristics produced by decommissioning activities and assess porpoise behaviour. During decommissioning, sound pressure spectral density levels in the frequency range 100 Hz to 48 kHz were 30-40 dB higher than baseline, with vessel presence being the main source of noise. The study detected small-scale (< 2 km) and short-term porpoise displacement during decommissioning, with porpoise occurrence increasing immediately after this. These findings can inform the consenting process for future decommissioning projects.

Bionic study of distance-azimuth discrimination of multi-scattered point objects in bat bio-sonar.

This paper presents a novel approach to enhance the discrimination capacity of multi-scattered point objects in bat bio-sonar. A broadband interferometer mathematical model is developed, incorporating both distance and azimuth information, to simulate the transmitted and received signals of bats. The Fourier transform is employed to simulate the preprocessing step of bat information for feature extraction. Furthermore, the bat bio-sonar model based on convolutional neural network (BS-CNN) is constructed to compensate for the limitations of conventional machine learning and CNN networks, including three strategies: Mix-up data enhancement, joint feature and hybrid atrous convolution module. The proposed BS-CNN model emulates the perceptual nerves of the bat brain for distance-azimuth discrimination and compares with four conventional classifiers to assess its discrimination efficacy. Experimental results demonstrate that the overall discrimination accuracy of the BS-CNN model is 93.4%, surpassing conventional CNN networks and machine learning methods by at least 5.9%. This improvement validates the efficacy of the BS-CNN bionic model in enhancing the discrimination accuracy in bat bio-sonar and offers valuable references for radar and sonar target classification.

A tradeoff evolution between acoustic fat bodies and skull muscles in toothed whales.

Toothed whales have developed specialized echolocation abilities that are crucial for underwater activities. Acoustic fat bodies, including the melon, extramandibular fat body, and intramandibular fat body, are vital for echolocation. This study explores the transcriptome of acoustic fat bodies in toothed whales, revealing some insight into their evolutionary origins and ecological significance. Comparative transcriptome analysis of acoustic fat bodies and related tissues in a harbor porpoise and a Pacific white-sided dolphin reveals that acoustic fat bodies possess characteristics of both muscle and adipose tissue, occupying an intermediate position. The melon and extramandibular fat body exhibit specific muscle-related functions, implying an evolutionary connection between acoustic fat bodies and muscle tissue. Furthermore, we suggested that the melon and extramandibular fat body originate from intramuscular adipose tissue, a component of white adipose tissue. The extramandibular fat body has been identified as an evolutionary homolog of the masseter muscle, supported by the specific expression of MYH16, a pivotal protein in masticatory muscles. The intramandibular fat body, located within the mandibular foramen, shows possibilities of the presence of several immune-related functions, likely due to its proximity to bone marrow. Furthermore, this study sheds light on leucine modification in the catabolic pathway, which leads to the accumulation of isovaleric acid in acoustic fat bodies. Swallowing without chewing, a major toothed whale feeding ecology adaptation, makes the masticatory muscle redundant and leads to the formation of the extramandibular fat body. We propose that the intramuscular fat enlargement in facial muscles, which influences acoustic fat body development, is potentially related to the substantial reorganization of head morphology in toothed whales during aquatic adaptation.

What determines the information update rate in echolocating bats.

The rate of sensory update is one of the most important parameters of any sensory system. The acquisition rate of most sensory systems is fixed and has been optimized by evolution to the needs of the animal. Echolocating bats have the ability to adjust their sensory update rate which is determined by the intervals between emissions - the inter-pulse intervals (IPI). The IPI is routinely adjusted, but the exact factors driving its regulation are unknown. We use on-board audio recordings to determine how four species of echolocating bats with different foraging strategies regulate their sensory update rate during commute flights. We reveal strong correlations between the IPI and various echolocation and movement parameters. Specifically, the update rate increases when the signals' peak-energy frequency and intensity increases while the update rate decreases when flight speed and altitude increases. We suggest that bats control their information update rate according to the behavioral mode they are engaged in, while always maintaining sensory continuity. Specifically, we suggest that bats apply two modes of attention during commute flights. Our data moreover suggests that bats emit echolocation signals at accurate intervals without the need for external feedback.

The distress context of social calls evokes a fear response in the bat Pipistrellus abramus.

Bats primarily use sound information, including echolocation, for social communication. Bats under stressful conditions, for example when confronted by a predator, will emit aggressive social calls. The presentation of aggressive social calls, including distress calls (DCs), is known to increase heart rate (fH), but how this change in fH is related to the bat's sound perception and how this evokes behaviors such as the fear response is unknown. Herein, we show that the perception of a distress context induces freezing behavior as a fear response in bats. We found that bats responded by freezing and displayed increased fH when they were presented with a conspecific donor bat in a distress situation evoked by gentle poking with a cotton swab. In addition, when we presented two types of auditory oddball paradigms with different probabilities of DCs and echolocation calls (ECs), the bats' fH increased when DCs were presented as deviant or control stimuli within standard ECs but did not increase when DCs were presented as standard stimuli. These results suggest that the situational context created by the frequency of sound presentation, rather than simply a single sound feature, induces fH increases and freezing as fear responses in bats.

Representation of frequency-modulated sweeps in the cochlear nucleus of the big brown bat.

The cochlear nucleus (CN) receives ipsilateral input from the auditory nerve and projects to other auditory brainstem nuclei. Little is known about CN processing of signals used for echolocation. This study recorded multiple unit activity in the CN of anesthetized big brown bats (Eptesicus fuscus) to ultrasonic frequency-modulated (FM) sweeps differing in sweep direction. FM up-sweeps evoke larger peak amplitudes at shorter onset latencies and with smaller amplitude-latency trading ratios than FM down-sweeps. Variability of onset latencies is in the tens of microsecond ranges, indicating sharp temporal precision in the CN for coding of FM signals.

Convergence in hearing-related genes between echolocating birds and mammals.

Echolocation, the detection of objects by means of sound waves, has evolved independently in diverse animals. Echolocators include not only mammals such as toothed whales and yangochiropteran and rhinolophoid bats but also Rousettus fruit bats, as well as two bird lineages, oilbirds and swiftlets. In whales and yangochiropteran and rhinolophoid bats, positive selection and molecular convergence has been documented in key hearing-related genes, such as prestin (SLC26A5), but few studies have examined these loci in other echolocators. Here, we examine patterns of selection and convergence in echolocation-related genes in echolocating birds and Rousettus bats. Fewer of these loci were under selection in Rousettus or birds compared with classically recognized echolocators, and elevated convergence (compared to outgroups) was not evident across this gene set. In certain genes, however, we detected convergent substitutions with potential functional relevance, including convergence between Rousettus and classic echolocators in prestin at a site known to affect hair cell electromotility. We also detected convergence between Yangochiroptera, Rhinolophidea, and oilbirds in TMC1, an important mechanosensory transduction channel in vertebrate hair cells, and observed an amino acid change at the same site within the pore domain. Our results suggest that although most proteins implicated in echolocation in specialized mammals may not have been recruited in birds or Rousettus fruit bats, certain hearing-related loci may have undergone convergent functional changes. Investigating adaptations in diverse echolocators will deepen our understanding of this unusual sensory modality.

Variations in echolocation click characteristics of finless porpoise in response to day/night and absence/presence of vessel noise.

Small odontocetes produce echolocation clicks to feed and navigate, making it an essential function for their survival. Recently, the effect of vessel noise on small odontocetes behavior has attracted attention owing to increase in vessel activities; however, the effects of the surrounding environmental factor, vessel noise, and day/night on echolocation click characteristics have not been well studied. Here, we examined the effects of vessel noise and day/night on variations in echolocation clicks and click trains parameters. Passive acoustic monitoring of on-axis echolocation clicks produced by free-ranging finless porpoises (Neophocaena asiaeorientalis sunameri) was performed at two sites in Japan, Seto Inland Sea and Mikawa Bay, in June-September 2021 and March-August 2022, using A-tag and SoundTrap 300HF. Generalized Linear Model was used to elucidate the effect of vessel noise, day/night, and surrounding environmental factors (water temperature, synthetic flow velocity, and noise level) on echolocation click and click train parameters. Echolocation click and click train parameters were strongly affected by day/night, whereas the absence/presence vessel noise did not exhibit statistically significant influence. Particularly, -3 dB bandwidth was wider, click duration was shorter, and inter-click intervals in a train were shorter at night, which may facilitate information processing at night, thereby compensating for the lack of visual information. The interaction between day/night and the absence/presence of vessel noise affected the source level of finless porpoises, with higher levels observed in the absence of vessel noise during the daytime compared to other conditions at the site with low vessel traffic. Overall, these results suggest that echolocation clicks by finless porpoise were likely to fluctuate to adapt with surrounding complex environmental conditions, especially day/night.

Connexin36 RNA Expression in the Cochlear Nucleus of the Echolocating Bat, Eptesicus fuscus.

PURPOSE: The echolocating bat is used as a model for studying the auditory nervous system because its specialized sensory capabilities arise from general mammalian auditory percepts such as pitch and sound source localization. These percepts are mediated by precise timing within neurons and networks of the lower auditory brainstem, where the gap junction protein Connexin36 (CX36) is expressed. Gap junctions and electrical synapses in the central nervous system are associated with fast transmission and synchronous patterns of firing within neuronal networks. The purpose of this study was to identify areas where CX36 was expressed in the bat cochlear nucleus to shed light on auditory brainstem networks in a hearing specialist animal model. METHODS: We investigated the distribution of CX36 RNA throughout the cochlear nucleus complex of the echolocating big brown bat, Eptesicus fuscus, using in situ hybridization. As a qualitative comparison, we visualized Gjd2 gene expression in the cochlear nucleus of transgenic CX36 reporter mice, species that hear ultrasound but do not echolocate. RESULTS: In both the bat and the mouse, CX36 is expressed in the anteroventral and in the dorsal cochlear nucleus, with more limited expression in the posteroventral cochlear nucleus. These results are generally consistent with previous work based on immunohistochemistry. CONCLUSION: Our data suggest that the anatomical substrate for CX36-mediated electrical neurotransmission is conserved in the mammalian CN across echolocating bats and non-echolocating mice.

Anatomy and homology of the caudal auricular muscles in greater short-nosed fruit bat (Cynopterus sphinx).

Bats can be phylogenetically classified into three major groups: pteropodids, rhinolophoids, and yangochiropterans. While rhinolophoids and yangochiropterans are capable of laryngeal echolocation, pteropodids lack this ability. Delicate ear movements are essential for echolocation behavior in bats with laryngeal echolocation. Caudal auricular muscles, especially the cervicoauricularis group, play a critical role in such ear movements. Previously, caudal auricular muscles were studied in three species of bats with laryngeal echolocation, but to our knowledge, there have been no studies on non-laryngeal echolocators, the pteropodids. Here, we describe the gross anatomy of the cervicoauricularis muscles and their innervation in Cynopterus sphinx by using diffusible iodine-based contrast-enhanced computed tomography and 3D reconstructions of immunohistochemically stained serial sections. A previous study on bats with laryngeal echolocation reported that rhinolophoids have four cervicoauricularis muscles and yangochiropterans have three. We observed three cervicoauricularis muscles in the pteropodid C. sphinx. The number of cervicoauricularis muscles and their innervation pattern were comparable to those of non-bat boreoeutherian mammals and yangochiropterans, suggesting that pteropodids, and yangochiropterans maintain the general condition of boreoeutherian mammals and that rhinolophoids have a derived condition. The unique nomenclature had been previously applied to the cervicoauricularis muscles of bats with laryngeal echolocation, but given the commonality between non-bat laurasiatherians and bats, with the exception of rhinolophoids, maintaining the conventional nomenclature (i.e., M. cervicoauricularis superficialis, M. cervicoauricularis medius, and M. cervicoauricularis profundus) is proposed for bats.

Background noise responding neurons in the inferior colliculus of the CF-FM bat, Hipposideros pratti.

The Lombard effect, referring to an involuntary rise in vocal intensity, is a widespread vertebrate mechanism that aims to maintain signal efficiency in response to ambient noise. Previous studies showed that the Lombard effect could be sufficiently implemented at subcortical levels and operated by continuously monitoring background noise, requiring some subcortical auditory sensitive neurons to have continuous responses to background noise. However, such neurons have not been well characterized. The inferior colliculus (IC) is a major auditory integration center under the auditory cortex and provides projections to the putative vocal pattern generator in the brainstem. Thus, it is reasonable to speculate that the IC is a likely auditory nucleus candidate having background noise responding neurons (BNR neurons). In the present study, we isolated 183 sound-sensitive IC neurons in a constant frequency-frequency modulation bat, Hipposideros pratti, and found that around 19% of these IC neurons are BNR neurons when stimulated with 70 dB SPL background white noise. Their firing rates in response to noise increased with increasing noise intensity and could be suppressed by sound stimulation. Furthermore, compared to neurons with similar best frequencies, the BNR neurons had smaller Q10-dB values and lower noise-induced minimal threshold change, indicating that BNR neurons received fewer inhibitory inputs. These results suggested that the BNR neurons are ideal candidates for collecting information about background noise. We proposed that the BNR neurons synapsed with neurons in vocal-pattern-generating networks in the brainstem and initiated the Lombard effect by a feed-forward loop.

Caudal auricular muscle variations and the evolution of echolocation behavior in pteropodid bats.

Among bats, rhinolophoids and yangochiropterans, but not pteropodids, exhibit laryngeal echolocation. Although Rousettus has been regarded as the only pteropodid capable of echolocation using tongue clicks, recent evidence suggests that other species of pteropodids are also capable of echolocation using wing clicks. Studies on laryngeal echolocators suggest that delicate ear movements are essential for the echolocation behavior of bats and that the cervicoauricularis muscles play a critical role in such ear movements. In this study, we observed the gross anatomy of cervicoauricularis muscles in three species of pteropodids (Cynopterus sphinx, Eonycteris spelaea, and Rousettus leschenaultii) to examine whether ear muscle anatomy varies among pteropodids with different echolocation types and between pteropodids and laryngeal echolocating bats. We found that M. cervicoauricularis profundus originates from the nuchal crest in tongue-click echolocators (R. leschenaultii) and from the midline in wing-click echolocators (C. sphinx and E. spelaea). In general, tongue-click echolocation using high click rates is considered to be more sophisticated in terms of sonar performance than wing-click echolocation. M. cervicoauricularis profundus originating from the nuchal crest (CPNC) is not common in non-bat laurasiatherian mammals, but can be found in laryngeal echolocating bats. As it pulls the ear pinna caudally in the horizontal plane and increases the access to sound, CPNC found in R. leschenaultii and laryngeal echolocating bats may be a key characteristic of the sophisticated active echolocation behavior of bats.

Extremely low seasonal prey capture efficiency in a deep-diving whale, the narwhal.

Successful foraging is essential for individuals to maintain the positive energy balance required for survival and reproduction. Yet, prey capture efficiency is poorly documented in marine apex predators, especially deep-diving mammals. We deployed acoustic tags and stomach temperature pills in summer to collect concurrent information on presumed foraging activity (through buzz detection) and successful prey captures (through drops in stomach temperature), providing estimates of feeding efficiency in narwhals. Compared to the daily number of buzzes (707 ± 368), the daily rate of feeding events was particularly low in summer (19.8 ± 8.9) and only 8-14% of the foraging dives were successful (i.e. with a detectable prey capture). This extremely low success rate resulted in a very low daily food consumption rate (less than 0.5% of body mass), suggesting that narwhals rely on body reserves accumulated in winter to sustain year-round activities. The expected changes or disappearance of their wintering habitats in response to climate change may therefore have severe fitness consequences for narwhal populations.

Oscillatory discharges in the auditory midbrain of the big brown bat contribute to coding of echo delay.

Subsequent to his breakthrough discovery of delay-tuned neurons in the bat's auditory midbrain and cortex, Albert Feng proposed that neural computations for echo delay involve intrinsic oscillatory discharges generated in the inferior colliculus (IC). To explore further the presence of these neural oscillations, we recorded multiple unit activity with a novel annular low impedance electrode from the IC of anesthetized big brown bats and Seba's short-tailed fruit bats. In both species, responses to tones, noise bursts, and FM sweeps contain long latency components, extending up to 60 ms post-stimulus onset, organized in periodic, oscillatory-like patterns at frequencies of 360-740 Hz. Latencies of this oscillatory activity resemble the wide distributions of single neuron response latencies in the IC. In big brown bats, oscillations lasting up to 30 ms after pulse onset emerge in response to single FM pulse-echo pairs, at particular pulse-echo delays. Oscillatory responses to pulses and evoked responses to echoes overlap extensively at short echo delays (5-7 ms), creating interference-like patterns. At longer echo delays, responses are separately evident to both pulses and echoes, with less overlap. These results extend Feng's reports of IC oscillations, and point to different processing mechanisms underlying perception of short vs long echo delays.

Echo feedback mediates noise-induced vocal modifications in flying bats.

Diverse animal taxa are capable of rapidly modifying vocalizations to mitigate interference from environmental noise. Echolocating bats, for example, must frequently perform sonar tasks in the presence of interfering sounds. Numerous studies have documented sound production flexibility in echolocating bats; however, it remains unknown whether noise-induced vocal modifications (NIVMs) mitigate interference effects on echoes or calls. In this study, we leverage echo level compensation behavior of echolocating bats to answer this question. Using a microphone array, we recorded echolocation calls of Hipposideros pratti trained to approach and land on a perch in the laboratory under quiet and noise conditions. We found that H. pratti exhibited echo level compensation behavior during approaching flights, which depended critically on distance to the landing perch. Broadcast noise delayed and affected the rate of echo level compensation in H. pratti. Moreover, H. pratti increased vocalization amplitude, i.e., exhibited the Lombard effect, while also adjusting call duration and bandwidth with increasing noise levels. Quantitative analyses of the data show that H. pratti relies on echo feedback, not vocal feedback, to adjust signals in the presence of noise. These findings provide compelling evidence that NIVMs in echolocating animals and non-echolocating animals operate through different mechanisms.

Transmitter and receiver of the low frequency horseshoe bat Rhinolophus paradoxolophus are functionally matched for fluttering target detection.

Flutter-detecting foragers require specific adaptations of the transmitter and the receiver of their echolocation systems to detect and evaluate flutter information in the echoes of potential prey. These adaptations include Doppler shift compensation (DSC), which keeps the echo frequency from targets ahead constant at a reference frequency (fref), and an auditory fovea in the cochlea, which results in foveal areas in the hearing system with many sharply tuned neurons with best frequencies near fref. So far, this functional match has been verified only for a very few key species, but is postulated for all flutter-detecting foragers. In this study we determined both, the transmitter and receiver properties within individuals of the Bourret's horseshoe bat (Rhinolophus paradoxolophus), an allometric outlier in the rhinolophid family. Here we show that the transmitter and receiver are functionally matched in a similar way as postulated for all flutter-detecting foragers. The performance of DSC, measured as the ability to keep the echo frequency constant at fref, had a precision similar to that found in other flutter-detecting foragers, and the audiogram showed the characteristic course with a minimum at fref. Furthermore, we show for a rhinolophid bat a variation over time of the coupled resting frequency and fref. Finally, we discuss the tight match between transmitter and receiver properties, which is guaranteed by the link between the foveal areas of the receiver and the audio-vocal control system for DSC.

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".

Brain-inspired sensorimotor echolocation system for confident landmark recognition.

A landmark is a familiar target in terms of the echoes that it can produce and is important for echolocation-based navigation by bats, robots, and blind humans. A brain-inspired system (BIS) achieves confident recognition, defined as classification to an arbitrarily small error probability (PE), by employing a voting process with an echo sequence. The BIS contains sensory neurons implemented with binary single-layer perceptrons trained to classify echo spectrograms with PE and generate excitatory and inhibitory votes in face neurons until a landmark-specific face neuron achieves recognition by reaching a confidence vote level (CVL). A discrete random step process models the vote count to show the recognition probability can achieve any desired accuracy by decreasing PE or increasing CVL. A hierarchical approach first classifies surface reflector and volume scatterer target categories and then uses that result to classify two subcategories that form four landmarks. The BIS models blind human echolocation to recognize four human-made and foliage landmarks by acquiring suitably sized and dense audible echo sequences. The sensorimotor BIS employs landmark-specific CVL values and a 2.7° view increment to acquire echo sequences that achieve zero-error recognition of each landmark independent of the initial view.