Plant classification from bat-like echolocation signals

Classification of plants according to their echoes is an elementary component of bat behavior that plays an important role in spatial orientation and food acquisition. Vegetation echoes are, however, highly complex stochastic signals: from an acoustical point of view, a plant can be thought of as a three-dimensional array of leaves reflecting the emitted bat call. The received echo is therefore a superposition of many reflections. In this work we suggest that the classification of these echoes might not be such a troublesome routine for bats as formerly thought. We present a rather simple approach to classifying signals from a large database of plant echoes that were created by ensonifying plants with a frequency-modulated bat-like ultrasonic pulse. Our algorithm uses the spectrogram of a single echo from which it only uses features that are undoubtedly accessible to bats. We used a standard machine learning algorithm (SVM) to automatically extract suitable linear combinations of time and frequency cues from the spectrograms such that classification with high accuracy is enabled. This demonstrates that ultrasonic echoes are highly informative about the species membership of an ensonified plant, and that this information can be extracted with rather simple, biologically plausible analysis. Thus, our findings provide a new explanatory basis for the poorly understood observed abilities of bats in classifying vegetation and other complex objects.     

Neural coding of echo-envelope disparities in echolocating bats

The effective use of echolocation requires not only measuring the delay between the emitted call and returning echo to estimate the distance of an ensonified object. To locate an object in azimuth and elevation, the bat’s auditory system must analyze the returning echoes in terms of their binaural properties, i.e., the echoes’ interaural intensity and time differences (IIDs and ITDs). The effectiveness of IIDs for echolocation is undisputed, but when bats ensonify complex objects, the temporal structure of echoes may facilitate the analysis of the echo envelope in terms of envelope ITDs. Using extracellular recordings from the auditory midbrain of the bat, Phyllostomus discolor, we found a population of neurons that are sensitive to envelope ITDs of echoes of their sonar calls. Moreover, the envelope-ITD sensitivity improved with increasing temporal fluctuations in the echo envelopes, a sonar parameter related to the spatial statistics of complex natural reflectors like vegetation. The data show that in bats envelope ITDs may be used not only to locate external, prey-generated rustling sounds but also in the context of echolocation. Specifically, the temporal fluctuations in the echo envelope, which are created when the sonar emission is reflected from a complex natural target, support ITD-mediated echolocation.     

Acoustic traits of bat-pollinated flowers compared to flowers of other pollination syndromes and their echo-based classification using convolutional neural networks

Bat-pollinated flowers have to attract their pollinators in absence of light and therefore some species developed specialized echoic floral parts. These parts are usually concave shaped and act like acoustic retroreflectors making the flowers acoustically conspicuous to the bats. Acoustic plant specializations only have been described for two bat-pollinated species in the Neotropics and one other bat-dependent plant in South East Asia. However, it remains unclear whether other bat-pollinated plant species also show acoustic adaptations. Moreover, acoustic traits have never been compared between bat-pollinated flowers and flowers belonging to other pollination syndromes. To investigate acoustic traits of bat-pollinated flowers we recorded a dataset of 32320 flower echoes, collected from 168 individual flowers belonging to 12 different species. 6 of these species were pollinated by bats and 6 species were pollinated by insects or hummingbirds. We analyzed the spectral target strength of the flowers and trained a convolutional neural network (CNN) on the spectrograms of the flower echoes. We found that bat-pollinated flowers have a significantly higher echo target strength, independent of their size, and differ in their morphology, specifically in the lower variance of their morphological features. We found that a good classification accuracy by our CNN (up to 84%) can be achieved with only one echo/spectrogram to classify the 12 different plant species, both bat-pollinated and otherwise, with bat-pollinated flowers being easier to classify. The higher classification performance of bat-pollinated flowers can be explained by the lower variance of their morphology.     

Object-oriented echo perception and cortical representation in echolocating bats

Echolocating bats can identify three-dimensional objects exclusively through the analysis of acoustic echoes of their ultrasonic emissions. However, objects of the same structure can differ in size, and the auditory system must achieve a size-invariant, normalized object representation for reliable object recognition. This study describes both the behavioral classification and the cortical neural representation of echoes of complex virtual objects that vary in object size. In a phantom-target playback experiment, it is shown that the bat Phyllostomus discolor spontaneously classifies most scaled versions of objects according to trained standards. This psychophysical performance is reflected in the electrophysiological responses of a population of cortical units that showed an object-size invariant response (14/109 units, 13%). These units respond preferentially to echoes from objects in which echo duration (encoding object depth) and echo amplitude (encoding object surface area) co-varies in a meaningful manner. These results indicate that at the level of the bat's auditory cortex, an object-oriented rather than a stimulus-parameter-oriented representation of echoes is achieved.     

Classification of insects by echolocating greater horseshoe bats

Summary Echolocating greater horseshoe bats (Rhinolophus ferrumequinum) detect insects by concentrating on the characteristic amplitude- and frequency modulation pattern fluttering insects impose on the returning echoes. This study shows that horseshoe bats can also further analyse insect echoes and thus recognize and categorize the kind of insect they are echolocating.Four greater horseshoe bats were trained in a twoalternative forced-choice procedure to choose the echo of one particular insect species turning its side towards the bat (Fig. 1). The bats were able to discriminate with over 90% correct choices between the reward-positive echo and the echoes of other insect species all fluttering with exactly the same wingbeat rate (Fig. 4).When the angular orientation of the reward-positive insect was changed (Fig. 2), the bats still preferred these unknown echoes over echoes from other insect species (Fig. 5) without any further training. Because the untrained bats did not show any prey preference, this indicates that the bats were able to perform an aspect-anglein-dependent classification of insects.Finally we tested what parameters in the echo were responsible for species recognition. It turned out that the bats especially used the small echo-modulations in between glints as a source of information (Fig. 7). Neither the amplitudenor the frequencymodulation of the echoes alone was sufficient for recognition of the insect species (Fig. 8). Bats performed a pattern recognition task based on complex computations of several acoustic parameters, an ability which might be termed cognitive.Abbreviations AM amplitude modulation - CF constant frequency - FM frequency modulation - S+ positive stimulus - S- negative stimulus     

On-board telemetry of emitted sounds from free-flying bats: compensation for velocity and distance stabilizes echo frequency and amplitude

To understand complex sensory-motor behavior related to object perception by echolocating bats, precise measurements are needed for echoes that bats actually listen to during flight. Recordings of echolocation broadcasts were made from flying bats with a miniature light-weight microphone and radio transmitter (Telemike) set at the position of the bat's ears and carried during flights to a landing point on a wall. Telemike recordings confirm that flying horseshoe bats (Rhinolophus ferrumequinum nippon) adjust the frequency of their sonar broadcasts to compensate for echo Doppler shifts. Returning constant frequency echoes were maintained at the bat's reference frequency +/-83 Hz during flight, indicating that the bats compensated for frequency changes with an accuracy equivalent to that at rest. The flying bats simultaneously compensate for increases in echo amplitude as target range becomes shorter. Flying bats thus receive echoes with both stabilized frequencies and stabilized amplitudes. Although it is widely understood that Doppler-shift frequency compensation facilitates detection of fluttering insects, approaches to a landing do not involve fluttering objects. Combined frequency and amplitude compensation may instead be for optimization of successive frequency modulated echoes for target range estimation to control approach and landing.     

Convergence of temporal and spectral information into acoustic images of complex sonar targets perceived by the echolocating bat,Eptesicus fuscus

1. FM echolocating bats (Eptesicus fuscus) were trained to discriminate between a two-component complex target and a one-component simple target simulated by electronically-returned echoes in a series of experiments that explore the composition of the image of the two-component target. In Experiment I, echoes for each target were presented sequentially, and the bats had to compare a stored image of one target with that of the other. The bats made errors when the range of the simple target corresponded to the range of either glint in the complex target, indicating that some trace of the parts of one image interfered with perception of the other image. In Experiment II, echoes were presented simultaneously as well as sequentially, permitting direct masking of echoes from one target to the other. Changes in echo amplitude produced shifts in apparent range whose pattern depended upon the mode of echo presentation. 2. Eptesicus perceives images of complex sonar targets that explicitly represent the location and spacing of discrete glints located at different ranges. The bat perceives the target's structure in terms of its range profile along a psychological range axis using a combination of echo delay and echo spectral representations that together resemble a spectrogram of the FM echoes. The image itself is expressed entirely along a range scale that is defined with reference to echo delay. Spectral information contributes to the image by providing estimates of the range separation of glints, but it is transformed into these estimates. 3. Perceived absolute range is encoded by the timing of neural discharges and is vulnerable to shifts caused by neural amplitude-latency trading, which was estimated at 13 to 18 microseconds per dB from N1 and N4 auditory evoked potentials in Eptesicus. Spectral cues representing the separation of glints within the target are transformed into estimates of delay separations before being incorporated into the image. However, because they are encoded by neural frequency tuning rather than the time-of-occurrence of neural discharges, the perceived range separation of glints in images is not vulnerable to amplitude-latency shifts. 4. The bat perceives an image that is displayed in the domain of time or range. The image receives no evident spectral contribution beyond what is transformed into delay estimates. Although the initial auditory representation of FM echoes is spectrogram-like, the time, frequency, and amplitude dimensions of the spectrogram appear to be compressed into an image that has only time and amplitude dimensions.(ABSTRACT TRUNCATED AT 400 WORDS)     

Size does not matter: size-invariant echo-acoustic object classification

Echolocating bats can not only extract spatial information from the auditory analysis of their ultrasonic emissions, they can also discriminate, classify and identify the three-dimensional shape of objects reflecting their emissions. Effective object recognition requires the segregation of size and shape information. Previous studies have shown that, like in visual object recognition, bats can transfer an echo-acoustic object discrimination task to objects of different size and that they spontaneously classify scaled versions of virtual echo-acoustic objects according to trained virtual-object standards. The current study aims to bridge the gap between these previous findings using a different class of real objects and a classification—instead of a discrimination paradigm. Echolocating bats (Phyllostomus discolor) were trained to classify an object as either a sphere or an hour-glass shaped object. The bats spontaneously generalised this classification to objects of the same shape. The generalisation cannot be explained based on similarities of the power spectra or temporal structures of the echo-acoustic object images and thus require dedicated neural mechanisms dealing with size-invariant echo-acoustic object analysis. Control experiments with human listeners classifying the echo-acoustic images of the objects confirm the universal validity of auditory size invariance. The current data thus corroborate and extend previous psychophysical evidence for sonar auditory-object normalisation and suggest that the underlying auditory mechanisms following the initial neural extraction of the echo-acoustic images in echolocating bats may be very similar in bats and humans.     

How Can Dolphins Recognize Fish According to Their Echoes? A Statistical Analysis of Fish Echoes

Echo-based object classification is a fundamental task of animals that use a biosonar system. Dolphins and porpoises should be able to rely on echoes to discriminate a predator from a prey or to select a desired prey from an undesired object. Many studies have shown that dolphins and porpoises can discriminate between objects according to their echoes. All of these studies however, used unnatural objects that can be easily characterized in human terminologies (e.g., metallic spheres, disks, cylinders). In this work, we collected real fish echoes from many angles of acquisition using a sonar system that mimics the emission properties of dolphins and porpoises. We then tested two alternative statistical approaches in classifying these echoes. Our results suggest that fish species can be classified according to echoes returning from porpoise- and dolphin-like signals. These results suggest how dolphins and porpoises can classify fish based on their echoes and provide some insight as to which features might enable the classification.     

Bats' avoidance of real and virtual objects: implications for the sonar coding of object size

Fast movement in complex environments requires the controlled evasion of obstacles. Sonar-based obstacle evasion involves analysing the acoustic features of object-echoes (e.g., echo amplitude) that correlate with this object's physical features (e.g., object size). Here, we investigated sonar-based obstacle evasion in bats emerging in groups from their day roost. Using video-recordings, we first show that the bats evaded a small real object (ultrasonic loudspeaker) despite the familiar flight situation. Secondly, we studied the sonar coding of object size by adding a larger virtual object. The virtual object echo was generated by real-time convolution of the bats’ calls with the acoustic impulse response of a large spherical disc and played from the loudspeaker. Contrary to the real object, the virtual object did not elicit evasive flight, despite the spectro-temporal similarity of real and virtual object echoes. Yet, their spatial echo features differ: virtual object echoes lack the spread of angles of incidence from which the echoes of large objects arrive at a bat's ears (sonar aperture). We hypothesise that this mismatch of spectro-temporal and spatial echo features caused the lack of virtual object evasion and suggest that the sonar aperture of object echoscapes contributes to the sonar coding of object size.     

A modeling approach to explain pulse design in bats

In this modeling study we wanted to find out why bats of the family Vespertilionidae (and probably also members of other families of bats) use pulses with a certain bandwidth and duration. Previous studies have only speculated on the function of bandwidth and pulse duration in bat echolocation or addressed this problem by assuming that bats optimize echolocation parameters to achieve very fine acuities in receiving single echoes. Here, we take a different approach by assuming that bats in nature rarely receive single echoes from each pulse emission, but rather many highly overlapping echoes. Some echolocation tasks require individual echoes to be separated to reconstruct reflection points in space. We used an established hearing model to investigate how the parameters bandwidth and pulse duration influence the separation of overlapping echoes. Our findings corroborate the following previously unknown or unsubstantiated facts: 1. Broadening the bandwidth improves the bat's lower resolution limit. 2. Increasing the sweep rate (defined by bandwidth and pulse duration) improves acuity of each extracted echo. 3. Decreasing the sweep rate improves the probability of frequency channels being activated. Since facts 2 and 3 affect sweep rate in an opposing fashion, an optimum sweep rate will exist, depending on the quality of the returning echoes and the requirements of the bat to improve acuity. The existence of an optimal sweep rate explains why bats are likely to use certain combinations of bandwidth and pulse duration to obtain such sweep rates.     

Dominant glint based prey localization in horseshoe bats: a possible strategy for noise rejection

Rhinolophidae or Horseshoe bats emit long and narrowband calls. Fluttering insect prey generates echoes in which amplitude and frequency shifts are present, i.e. glints. These glints are reliable cues about the presence of prey and also encode certain properties of the prey. In this paper, we propose that these glints, i.e. the dominant glints, are also reliable signals upon which to base prey localization. In contrast to the spectral cues used by many other bats, the localization cues in Rhinolophidae are most likely provided by self-induced amplitude modulations generated by pinnae movement. Amplitude variations in the echo not introduced by the moving pinnae can be considered as noise interfering with the localization process. The amplitude of the dominant glints is very stable. Therefore, these parts of the echoes contain very little noise. However, using only the dominant glints potentially comes at a cost. Depending on the flutter rate of the insect, a limited number of dominant glints will be present in each echo giving the bat a limited number of sample points on which to base localization. We evaluate the feasibility of a strategy under which Rhinolophidae use only dominant glints. We use a computational model of the echolocation task faced by Rhinolophidae. Our model includes the spatial filtering of the echoes by the morphology of the sonar apparatus of Rhinolophus rouxii as well as the amplitude modulations introduced by pinnae movements. Using this model, we evaluate whether the dominant glints provide Rhinolophidae with enough information to perform localization. Our simulations show that Rhinolophidae can use dominant glints in the echoes as carriers for self-induced amplitude modulations serving as localization cues. In particular, it is shown that the reduction in noise achieved by using only the dominant glints outweighs the information loss that occurs by sampling the echo.     

Efficient encoding of spectrotemporal information for bat echolocation

In most animals, natural stimuli are characterized by a high degree of redundancy, limiting the ensemble of ecologically valid stimuli to a significantly reduced subspace of the representation space. Neural encodings can exploit this redundancy and increase sensing efficiency by generating low-dimensional representations that retain all information essential to support behavior. In this study, we investigate whether such an efficient encoding can be found to support a broad range of echolocation tasks in bats. Starting from an ensemble of echo signals collected with a biomimetic sonar system in natural indoor and outdoor environments, we use independent component analysis to derive a low-dimensional encoding of the output of a cochlear model. We show that this compressive encoding retains all essential information. To this end, we simulate a range of psycho-acoustic experiments with bats. In these simulations, we train a set of neural networks to use the encoded echoes as input while performing the experiments. The results show that the neural networks’ performance is at least as good as that of the bats. We conclude that our results indicate that efficient encoding of echo information is feasible and, given its many advantages, very likely to be employed by bats. Previous studies have demonstrated that low-dimensional encodings allow for task resolution at a relatively high level. In contrast to previous work in this area, we show that high performance can also be achieved when low-dimensional filters are derived from a data set of realistic echo signals, not tailored to specific experimental conditions.     

What a Plant Sounds Like: The Statistics of Vegetation Echoes as Received by Echolocating Bats

A critical step on the way to understanding a sensory system is the analysis of the input it receives. In this work we examine the statistics of natural complex echoes, focusing on vegetation echoes. Vegetation echoes constitute a major part of the sensory world of more than 800 species of echolocating bats and play an important role in several of their daily tasks. Our statistical analysis is based on a large collection of plant echoes acquired by a biomimetic sonar system. We explore the relation between the physical world (the structure of the plant) and the characteristics of its echo. Finally, we complete the story by analyzing the effect of the sensory processing of both the echolocation and the auditory systems on the echoes and interpret them in the light of information maximization. The echoes of all different plant species we examined share a surprisingly robust pattern that was also reproduced by a simple Poisson model of the spatial reflector arrangement. The fine differences observed between the echoes of different plant species can be explained by the spatial characteristics of the plants. The bat's emitted signal enhances the most informative spatial frequency range where the species-specific information is large. The auditory system filtering affects the echoes in a similar way, thus enhancing the most informative spatial frequency range even more. These findings suggest how the bat's sensory system could have evolved to deal with complex natural echoes.     

Different Auditory Feedback Control for Echolocation and Communication in Horseshoe Bats

Auditory feedback from the animal''s own voice is essential during bat echolocation: to optimize signal detection, bats continuously adjust various call parameters in response to changing echo signals. Auditory feedback seems also necessary for controlling many bat communication calls, although it remains unclear how auditory feedback control differs in echolocation and communication. We tackled this question by analyzing echolocation and communication in greater horseshoe bats, whose echolocation pulses are dominated by a constant frequency component that matches the frequency range they hear best. To maintain echoes within this “auditory fovea”, horseshoe bats constantly adjust their echolocation call frequency depending on the frequency of the returning echo signal. This Doppler-shift compensation (DSC) behavior represents one of the most precise forms of sensory-motor feedback known. We examined the variability of echolocation pulses emitted at rest (resting frequencies, RFs) and one type of communication signal which resembles an echolocation pulse but is much shorter (short constant frequency communication calls, SCFs) and produced only during social interactions. We found that while RFs varied from day to day, corroborating earlier studies in other constant frequency bats, SCF-frequencies remained unchanged. In addition, RFs overlapped for some bats whereas SCF-frequencies were always distinctly different. This indicates that auditory feedback during echolocation changed with varying RFs but remained constant or may have been absent during emission of SCF calls for communication. This fundamentally different feedback mechanism for echolocation and communication may have enabled these bats to use SCF calls for individual recognition whereas they adjusted RF calls to accommodate the daily shifts of their auditory fovea.     

Adaptive mechanisms underlying the bat biosonar behavior

For survival, bats of the suborder Microchiropetra emit intense ultrasonic pulses and analyze the weak returning echoes to extract the direction, distance, velocity, size, and shape of the prey. Although these bats and other mammals share the common layout of the auditory pathway and sound coding mechanism, they have highly developed auditory systems to process biologically relevant pulses at the expense of a reduced visual system. During this active biosonar behavior, they progressively shorten the pulse duration, decrease the amplitude and pulse-echo gap as they search, approach and finally intercept the prey. Presumably, these changes in multiple pulse parameters throughout the entire course of hunting enable them to extract maximal information about localized prey from the returning echoes. To hunt successfully, the auditory system of these bats must be less sensitive to intense emitted pulses but highly sensitive to weak returning echoes. They also need to recognize and differentiate the echoes of their emitted pulses from echoes of pulses emitted by other conspecifics. Past studies have shown the following mechanical and neural adaptive mechanisms underlying the successful bat biosonar behavior: (1) Forward orienting and highly mobile pinnae for effective scanning, signal reception, sound pressure transformation and mobile auditory sensitivity; (2) Avoiding and detecting moving targets more successfully than stationary ones; (3) Coordinated activity of highly developed laryngeal and middle ear muscles during pulse emission and reception; (4) Mechanical and neural attenuation of intense emitted pulses to prepare for better reception of weak returning echoes; (5) Increasing pulse repetition rate to improve multiple-parametric selectivity to echoes; (6) Dynamic variation of duration selectivity and recovery cycle of auditory neurons with hunting phase for better echo analysis; (7) Maximal multiple-parametric selectivity to expected echoes returning within a time window after pulse emission; (8) Pulse-echo delaysensitive neurons in higher auditory centers for echo ranging; (9) Corticofugal modulation to improve on-going multiple-parametric signal processing and reorganize signal representation, and (10) A large area of the superior colliculus, pontine nuclei and cerebellum that is sensitive to sound for sensori-motor integration. All these adaptive mechanisms facilitate the bat to effectively extract prey features for successful hunting.     

Fine control of call frequency by horseshoe bats

The auditory system of horseshoe bats is narrowly tuned to the sound of their own echoes. During flight these bats continuously adjust the frequency of their echolocation calls to compensate for Doppler-effects in the returning echo. Horseshoe bats can accurately compensate for changes in echo frequency up to 5 kHz, but they do so through a sequence of small, temporally-independent, step changes in call frequency. The relationship between an echo's frequency and its subsequent impact on the frequency of the very next call is fundamental to how Doppler-shift compensation behavior works. We analyzed how horseshoe bats control call frequency by measuring the changes occurring between many successive pairs of calls during Doppler-shift compensation and relating the magnitude of these changes to the frequency of each intervening echo. The results indicate that Doppler-shift compensation is mediated by a pair of (echo)frequency-specific sigmoidal functions characterized by a threshold, a slope, and an upper limit to the maximum change in frequency that may occur between successive calls. The exact values of these parameters necessarily reflect properties of the underlying neural circuitry of Doppler-shift compensation and the motor control of vocalization, and provide insight into how neural feedback can accommodate the need for speed without sacrificing stability.     

Generalization of 'same-different' classification abilities in bottlenosed dolphins

Two bottlenosed dolphins taught to classify pairs of three-dimensional objects as either same or different were tested with novel stimulus sets to determine how well their classification abilities would generalize. Both dolphins were immediately able to classify novel pairs of planar objects, differing only in shape, as same or different. When tested on sets of three objects consisting of either all different objects or of two identical objects and one different object, both dolphins proved to be able to classify 'all different' sets as different and 'not all different' sets as same, at levels significantly above chance. These data suggest that dolphins can use knowledge about similarity-based classification strategies gained from previous training to perform successfully in a variety of novel same-different classification tasks. Visual classificatory abilities of dolphins appear to be comparable to those that have been demonstrated in primates.     

Keeping returns optimal: gain control exerted through sensitivity adjustments in the harbour porpoise auditory system

Animals that use echolocation (biosonar) listen to acoustic signals with a large range of intensities, because echo levels vary with the fourth power of the animal's distance to the target. In man-made sonar, engineers apply automatic gain control to stabilize the echo energy levels, thereby rendering them independent of distance to the target. Both toothed whales and bats vary the level of their echolocation clicks to compensate for the distance-related energy loss. By monitoring the auditory brainstem response (ABR) during a psychophysical task, we found that a harbour porpoise (Phocoena phocoena), in addition to adjusting the sound level of the outgoing signals up to 5.4 dB, also reduces its ABR threshold by 6 dB when the target distance doubles. This self-induced threshold shift increases the dynamic range of the biosonar system and compensates for half of the variation of energy that is caused by changes in the distance to the target. In combination with an increased source level as a function of target range, this helps the porpoise to maintain a stable echo-evoked ABR amplitude irrespective of target range, and is therefore probably an important tool enabling porpoises to efficiently analyse and classify received echoes.