Click Signals Research Articles

WaveNet-based approximation of a cochlear filtering and hair cell transduction model

Computational auditory models are important tools for gaining new insights into hearing mechanisms, and they can provide a foundation for bio-inspired speech and audio processing algorithms. However, accurate models often entail an immense computational effort, rendering their application unfeasible if quick execution is required. This paper presents a WaveNet-based approximation of the normal-hearing cochlear filtering and inner hair cell (IHC) transduction stages of a widely used auditory model [Zilany and Bruce (2006). J. Acoust. Soc. Am. 120(3), 1446–1466]. The WaveNet model was trained and optimized using a large dataset of clean speech, noisy speech, and music for a wide range of sound pressure levels (SPLs) and characteristic frequencies between 125 Hz and 8 kHz. The model was evaluated with unseen (noisy) speech, music signals, sine tones, and click signals at SPLs between 30 and 100 dB. It provides accurate predictions of the IHC receptor potentials for a given input stimulus and allows an efficient execution with processing times up to 250 times lower compared to an already optimized reference implementation of the original auditory model. The WaveNet model is fully differentiable, thus, allowing its application in the context of deep-learning-based speech and audio enhancement algorithms.

Acoustic feeding responses using marine chemoattractants in plant-based diets for naive and non-naive Litopenaeus vannamei

The inclusion of chemoattractants has been investigated to improve the attractability and palatability of plant-based shrimp diets. A more recent approach utilizes passive acoustic monitoring (PAM) to record the sounds (clicks) emitted by shrimp during feeding activity allowing the identification of behavioral responses related to the diets. In the present study we automatically identified click signals in PAM recordings at sampling frequency of 192 kHz. This methodology was applied to evaluate the efficiency of three chemoattractants (krill meal, squid meal and fish hydrolysate) added to a basal all-plant diet on the acoustic feeding behavior and food consumption of naive (without acclimation) and non-naive (acclimated to the diets) Litopenaeus vannamei. Five diets were used: basal all-plant diet (AP, negative control) and its modifications to contain 20 g/Kg of krill meal (AP+KM), 20 g/Kg of squid meal (AP+SM) and 40 g/Kg of fish hydrolysate (AP+FH); a diet with 120 g/Kg fishmeal (FM) was used as a positive control. Ten shrimp (6.8 ± 0.6 g; mean ± SD) were stocked per aquarium (70 L) in a clear water recirculation system (30 aquaria, five diets with six replicates). The food consumption and acoustic activity were evaluated simultaneously in each aquarium over a 30-min period during two consecutive days for each trial, naive and non-naive shrimp. Strong positive relationships (r = 0.78–0.81; P < 0.001) were found between food consumption and total number of clicks emitted by naive and non-naive shrimp. Overall, the inclusion of the chemoattractants to the plant-based diet had a positive effect improving acoustic feeding behavior and food consumption for naive and non-naive L. vannamei. In particular, the response of non-naive shrimp to chemoattractants was markedly increased, and the highest clicking activity was attained by using krill meal (AP+KM), followed by fish hydrolysate (AP+FH). The new analytical method proved to be a useful alternative approach for studying shrimp feeding behavior, also indicating the potential of PAM to forecast feed intake and improve efficiency of feed management.

A dolphin-inspired compact sonar for underwater acoustic imaging

Underwater imaging sonars are widely used for oceanic exploration but are bulky and expensive for some applications. The sonar system of dolphins, which uses sound pulses called clicks to investigate their environment, offers superior shape discrimination capability compared to human-derived imaging sonars of similar size and frequency. In order to gain better understanding of dolphin sonar imaging, we train a dolphin to acoustically interrogate certain objects and match them visually. We record the echoes the dolphin receives and are able to extract object shape information from these recordings. We find that infusing prior information into the processing, specifically the sparsity of the shapes, yields a clearer interpretation of the echoes than conventional signal processing. We subsequently develop a biomimetic sonar system that combines sparsity-aware signal processing with high-frequency broadband click signals similar to that of dolphins, emitted by an array of transmitters. Our findings offer insights and tools towards compact higher resolution sonar imaging technologies.

Acoustic activity of Litopenaeus vannamei fed pelleted and extruded diets

Passive acoustics monitoring (PAM) is an exceptional tool to analyze the feeding behavior of shrimp, as they emit a “click” sound by the mandibular occlusion. Hence, acoustic demand-feeding systems based on the click signals energy have been applied to improve feeding management in shrimp farming. These acoustic signals may be affected by the texture properties of the foods pellets, which are modified by the manufacturing process. The aim of the present study was to evaluate the acoustic activity of Litopenaeus vannamei fed pelleted and extruded diets. Two commercial diets were used, manufactured by extrusion and pelleting processing with similar formulation (36% crude protein) and pellet diameter. Acoustic evaluation was performed with these diets offered “as is” dry and previously soaked in water for one and two hours in a 2 × 3 factorial arrangement. The pellet hardness and moisture content were also measured. The acoustic trial was performed using omnidirectional hydrophones inside acoustic foam anechoic chambers. Three fastened shrimp (10 g) were placed inside acoustic chambers and audio recordings started (30 min) when food (0.5 g) was offered. The texture of diets was closely related to the acoustic intensity (Root Mean Square Amplitude - RMS) produced by L. vannamei during the feeding activity. The greater hardness of the extruded diet resulted in higher RMS than pelleted food when offered dry, which could be an advantage for the sound detection of feeding activity. The lower RMS observed when both diets were soaked suggests that softer texture pellets can be eventually ingested without the need of mandibles occlusion, and consequently without clicking sound emission. Additionally, unusual clicks train signals were first reported when shrimp fed soaked diets, probably due to mandibles rubbing. The relationship between textural properties and acoustical responses of foods should gain more attention, especially considering the potential applications on automatic acoustic feedback systems in shrimp farming.

Signal processing for biological transients emitted by dolphins and snapping shrimp

In temperate and tropical waters, the dominant source of biological noise in shallow bays, harbors and inlets is snapping shrimp, which can adversely affect the performance of high-frequency sonar systems. The modified method for passive ranging by wavefront curvature enables the source position and source level of individual snaps to be measured in situ. The results are presented for a sequence of 1000 snaps. The principal habitat of the local snapping shrimp population coincides with a narrow wharf, which is 120 m in length. Next, this passive ranging method is used to localize free-ranging dolphins in their natural habitat. The peak-to-peak sound pressure levels of the biosonar pulses (clicks) emitted by Indo-Pacific dolphins (Tursiops aduncus) are measured for source ranges from 30 m to 300 m, even when the dolphins are echolocating at the same time. The waveforms and signal properties of the click signals are presented, along with their effectiveness for probing the shallow-water environment. The bistatic scattering impulse responses of the sea floor, sea surface and scatterers in the water volume are shown. Finally, consideration is given to the contexts in which snaps and clicks are made.

An extensible approach for real-time bidding with model-free reinforcement learning

In this paper, we propose an extensible framework for model-free reinforcement learning (RL) for real-time bidding (RTB) in display advertising. This framework can be applied into both simple environments and extend to the comprehensive environment that the DSP bids for multiple advertisers at the same time. To process new information that is collected via real-time interaction with the environment, an extensible model is first introduced, which is based on the distribution of the recharging probability. Substantial effort is expended to alleviate the problem of the sparsity of the click signal with the reward function. The proposed scheme has high feasibility and can address dynamic environments in contrast to prior works, which assumed that the distribution of the feature vectors and the dealing price were already known. Furthermore, a fund-recharging mechanism is introduced for transforming the RTB model into an endless task, which allows the policy to be optimized in a farsighted rather than a myopic manner. Illustrative experiments on both the small- and large-scale real datasets demonstrate the state-of-the-art performance of the proposed framework for the issue of interest.

Analysis of echolocation click signals of Indo-Pacific bottlenose dolphin (Tursiops aduncus) in Jeju Island

Analysis of echolocation click signals of Indo-Pacific bottlenose dolphin (Tursiops aduncus) in Jeju Island

Human Ability to Judge Relative Size and Lateral Position of a Sound Reflecting Board Using Click Signals: Influence of Source Position and Click Properties

In human echolocation, acoustic signals provide information enabling people to perceive the physical environment. Little is known about the relation between properties of click signals used for echolocation purposes and the effect of source position on the resulting echolocation performance. In this work, echolocation performance is studied for two different echolocation tests: board size identification and board lateralization. The used artificial source is placed in front of the mouth and in front of the chest. Three different synthesized test signals are used: two narrow band signals and one broadband click signal. Impulse responses were recorded in an anechoic room for each individual participant, and listening tests were carried out afterwards by playing post-processed individualized signals using headphones. Results show better performance with the sound source at the chest position in both experiments, with the difference in performance most prominent in the board size identification test. In the board lateralization test, the broadband click results in eminent better performance over narrow band click signals. This suggest that broadband and chest positioned signals seem favorable over narrow band and mouth positioned signals

Acoustic Properties of Tongue Clicks used for Human Echolocation

In active human echolocation, it is common for visually impaired persons to produce tongue click signals for extracting information from the physical environment around them. The purpose of this paper is to provide a comprehensive dataset of original detailed experimental data on human tongue clicks. The methods used for analysis are described in order to encourage further work to complement the data, which is available online and open for contributions. Series of six different click signals have been recorded from blind persons in an anechoic room. The different click signals all resemble damped oscillations and have in common that the dominant frequency content is located between 1,000 Hz and 8,000 Hz. The standard deviation in click series of the broadband click sound level is between 2.2 and 4.3 dB. The total duration of the clicks, which is mainly determined by the drop-time, ranges between 2.0 ms and 10.0 ms, for a 20 dB decay range. Average standard deviations of 57% in rise-time and 32% in drop-time have been found. The horizontal and vertical directivity of the tongue clicks show a close agreement with literature data from speech signals, but click signals are generally more directive. The minimum and maximum deviations in the 1/3 octave bands from 1,250 Hz to 4,000 Hz are less than 6 dB, for 0 to 120 deg. azimuth, and for –27 to 127 deg. elevation.

Bio‐inspired signal detection mechanism for tongue click waveform used in human echolocation

Human echolocation is the ability of an individual (which is often a blind person) to use his/her signal such as sound from tongue clicks to perceive the surrounding. Basically this requires the person to listen and analyse to the return echo of the tongue clicks. The main characteristics of the tongue click signal waveform have been reported, however the fundamental principle on a person's ability to identify his/her own signal is still vague. The possible detection mechanism of the tongue click signal waveform used in human echolocation technique is discussed and imitated it as artificial detection system. The proposed mechanism which is based on human hearing process in synthesising the signal illustrates that the detection performance is improved as compared to the detection performance by the traditional matched filtering technique. The findings of this Letter create new potential for the development of any artificial human echolocator system, sensor systems like radar and sonar as well as applications inspired by human echolocation miracles.

Dolphin target detection processes investigation using finite element model

Dolphins and porpoises use their sophisticated biosonar systems for targets detection, within a range of a few meters to about 200 m, there is not a better sonar on the planet. In this study, the high resolution computer tomography (CT) scan data were used to create the detecting click signal propagation models of Atlantic bottlenose dolphins (Tursiops truncatus) and harbor porpoise (Phocoena phocoena). The finite element methods (FEM) were used to simulate the processes of the clicks emitted from phonic lips and transmit to the water through animals’ heads. The biosonar beam forming in the nearfield and farfield including the amplitude contours were determined and compared to the prior measurement results. There were no evidences of convergence in the farfield, which were consistent with measurement results for Tursiops truncatus. Additionally, in a cross-modal matching experiments with Tursiops, we found that the accuracy of the successive match was significantly different when the following subjects with same shape were used (water-filled PVC pipes, air-filled PVC pipes, foam ball array, and PVC pipes wrapped by foam) the results of FE model shed some light on the reasons why the animal has significant difference in performances when detecting different targets in the experiments.

Transmission beam characteristics of a Risso's dolphin (Grampus griseus).

The echolocation system of the Risso's dolphin (Grampus griseus) remains poorly studied compared to other odontocete species. In this study, echolocation signals were recorded from a stationary Risso's dolphin with an array of 16 hydrophones and the two-dimensional beam shape was explored using frequency-dependent amplitude plots. Click source parameters were similar to those already described for this species. Centroid frequency of click signals increased with increasing sound pressure level, while the beamwidth decreased with increasing center frequency. Analysis revealed primarily single-lobed, and occasionally vertically dual-lobed, beam shapes. Overall beam directivity was found to be greater than that of the harbor porpoise, bottlenose dolphin, and a false killer whale. The relationship between frequency content, beam directivity, and head size for this Risso's dolphin deviated from the trend described for other species. These are the first reported measurements of echolocation beam shape and directivity in G. griseus.

Sound signatures and production mechanisms of three species of pipefishes (Family: Syngnathidae)

Background. Syngnathid fishes produce three kinds of sounds, named click, growl and purr. These sounds are generated by different mechanisms to give a consistent signal pattern or signature which is believed to play a role in intraspecific and interspecific communication. Commonly known sounds are produced when the fish feeds (click, purr) or is under duress (growl). While there are more acoustic studies on seahorses, pipefishes have not received much attention. Here we document the differences in feeding click signals between three species of pipefishes and relate them to cranial morphology and kinesis, or the sound-producing mechanism.Methods. The feeding clicks of two species of freshwater pipefishes, Doryichthys martensii and Doryichthys deokhathoides and one species of estuarine pipefish, Syngnathoides biaculeatus, were recorded by a hydrophone in acoustic dampened tanks. The acoustic signals were analysed using time-scale distribution (or scalogram) based on wavelet transform. A detailed time-varying analysis of the spectral contents of the localized acoustic signal was obtained by jointly interpreting the oscillogram, scalogram and power spectrum. The heads of both Doryichthys species were prepared for microtomographical scans which were analysed using a 3D imaging software. Additionally, the cranial bones of all three species were examined using a clearing and double-staining method for histological studies.Results. The sound characteristics of the feeding click of the pipefish is species-specific, appearing to be dependent on three bones: the supraoccipital, 1st postcranial plate and 2nd postcranial plate. The sounds are generated when the head of the Dorichthyes pipefishes flexes backward during the feeding strike, as the supraoccipital slides backwards, striking and pushing the 1st postcranial plate against (and striking) the 2nd postcranial plate. In the Syngnathoides pipefish, in the absence of the 1st postcranial plate, the supraoccipital rubs against the 2nd postcranial plate twice as it is pulled backward and released on the return. Cranial morphology and kinesis produce acoustic signals consistent with the bone strikes that produce sharp energy spikes (discrete or merged), or stridulations between bones that produce repeated or multimodal sinusoidal waveforms.Discussion. The variable structure of the sound-producing mechanism explains the unique acoustic signatures of the three species of pipefish. The differences in cranial bone morphology, cranial kinesis and acoustic signatures among pipefishes (and seahorses) could be attributed to independent evolution within the Syngnathidae, which warrants further investigation.

Echolocation beam shape of the Risso's dolphin (Grampus griseus)

Studies have shown that echolocation signals of some odontocete species are projected in both single and sometimes vertically dual-lobed beam shapes. In this study, the echolocation beam of a Risso's dolphin (Grampus griseus) was measured from a captive individual ensonifying an underwater target. Clicks were recorded with an array of 16 hydrophones and the two dimensional beam shape was explored using frequency-dependent amplitude plots. Click source parameters were comparable to those already described for this species, while analysis revealed primarily single-lobed, and occasionally vertically dual-lobed, beam shapes. Center frequencies of click signals increased with increasing sound pressure level, while the −3 dB beam radius decreased with increasing center frequency. This study is the first to measure the beam shape of echolocation signals in G. griseus, which exhibits forms similar to those found in the bottlenose dolphin and false killer whale

Towards high performance communication using intracortical brain-computer interfaces

Towards high performance communication using intracortical brain-computer interfaces

Location- and Query-Aware Modeling of Browsing and Click Behavior in Sponsored Search

An online advertisement’s clickthrough rate provides a fundamental measure of its quality, which is widely used in ad selection strategies. Unfortunately, ads placed in contexts where they are rarely viewed—or where users are unlikely to be interested in commercial results—may receive few clicks regardless of their quality. In this article, we model the variability of a user’s browsing behavior for the purpose of click analysis and prediction in sponsored search. Our model incorporates several important contextual factors that influence ad clickthrough rates, including the user’s query and ad placement on search engine result pages. We formally model these factors with respect to the list of ads displayed on a result page, the probability that the user will initiate browsing of this list, and the persistence of the user in browsing the list. We incorporate these factors into existing click models by augmenting them with appropriate query and location biases. Using expectation maximization, we learn the parameters of these augmented models from click signals recorded in the logs of a commercial search engine. To evaluate the performance of the models and to compare them with state-of-the-art performance, we apply standard evaluation metrics, including log-likelihood and perplexity. Our evaluation results indicate that, through the incorporation of query and location biases, significant improvements can be achieved in predicting browsing and click behavior in sponsored search. In addition, we explore the extent to which these biases actually reflect varying behavioral patterns. Our observations confirm that correlations exist between the biases and user search behavior.

Using a coherent hydrophone array for observing sperm whale range, classification, and shallow-water dive profiles

Sperm whales in the New England continental shelf and slope were passively localized, in both range and bearing, and classified using a single low-frequency (<2500 Hz), densely sampled, towed horizontal coherent hydrophone array system. Whale bearings were estimated using time-domain beamforming that provided high coherent array gain in sperm whale click signal-to-noise ratio. Whale ranges from the receiver array center were estimated using the moving array triangulation technique from a sequence of whale bearing measurements. Multiple concurrently vocalizing sperm whales, in the far-field of the horizontal receiver array, were distinguished and classified based on their horizontal spatial locations and the inter-pulse intervals of their vocalized click signals. The dive profile was estimated for a sperm whale in the shallow waters of the Gulf of Maine with 160 m water-column depth located close to the array's near-field where depth estimation was feasible by employing time difference of arrival of the direct and multiply reflected click signals received on the horizontal array. By accounting for transmission loss modeled using an ocean waveguide-acoustic propagation model, the sperm whale detection range was found to exceed 60 km in low to moderate sea state conditions after coherent array processing.

Using a coherent hydrophone array for observing sperm whale range, classification, and shallow-water dive profiles.

Sperm whales in the New England continental shelf and slope were passively localized, in both range and bearing, and classified using a single low-frequency (<2500 Hz), densely sampled, towed horizontal coherent hydrophone array system. Whale bearings were estimated using time-domain beamforming that provided high coherent array gain in sperm whale click signal-to-noise ratio. Whale ranges from the receiver array center were estimated using the moving array triangulation technique from a sequence of whale bearing measurements. Multiple concurrently vocalizing sperm whales, in the far-field of the horizontal receiver array, were distinguished and classified based on their horizontal spatial locations and the inter-pulse intervals of their vocalized click signals. The dive profile was estimated for a sperm whale in the shallow waters of the Gulf of Maine with 160 m water-column depth located close to the array's near-field where depth estimation was feasible by employing time difference of arrival of the direct and multiply reflected click signals received on the horizontal array. By accounting for transmission loss modeled using an ocean waveguide-acoustic propagation model, the sperm whale detection range was found to exceed 60 km in low to moderate sea state conditions after coherent array processing.

Release from auditory masking with complex signals and complex noise in the bottlenose dolphin (Tursiops truncatus)

Dolphins are social animals that rely heavily on passive and active acoustics for communication, navigation, foraging, and detecting predators. Auditory masking, from both natural and anthropogenic noise sources, may adversely affect these fitness-related capabilities. The dolphin's ability to detect a variety of complex signals (both dolphin phonations and tonal signals) masked by Gaussian, comodulated, snapping shrimp, and ice squeaks noise was tested. Detection thresholds were measured using a go/no-go adaptive staircase procedure. Masking patterns were similar for all signals (whistles, burst-pulse, and pure tones) except for click signals. Masking from ice squeaks resulted in the largest masked thresholds, while snapping shrimp and comodulated noise resulted in a release from masking relative to thresholds from Gaussian noise. Click signals were most difficult to detect when masked by snapping shrimp. Recognition thresholds were estimated for whistle-like signals using a cross-modal, matching-to-sample procedure. Recognition thresholds were on average 4 dB greater than detection thresholds for all noise types. The auditory mechanisms governing the results are discussed. [Work supported by the ONR.]

Characterization of Yellow Seahorse Hippocampus kuda feeding click sound signals in a laboratory environment: an application of probability density function and power spectral density analyses

Do the sounds generated by different-sized fish of different sexes differ from each other in temporal, spectral or intensity patterns? Such differences would enable the development of passive acoustic techniques to locate seahorses in open water bodies. On the basis of this perspective, we characterize and present in this study Yellow Seahorse Hippocampus kuda (Syngnathidae) feeding click sounds, recorded in a controlled laboratory environment. The characterization involved analysis of the “probability density function” (PDF) and the “power spectral density” (PSD) of the seahorse feeding click sound signals of different sizes and sexes. The PDF describes the general distribution of the magnitude of a random process, and such analysis, using recorded seahorse clicks, points towards a multimodal statistical distribution, which is the existence of more than one fitting component for the majority of seahorse click signals above the tank ambient noise level. This fact has been appraised towards the involvement of more than one process in the seahorse click signal generation mechanism. However, lack of prior knowledge about the individual PDF components of the data leads towards the mismatch between the data of PDFs. The PDF gives no information on the time and frequency content of the processes as provided by the PSD. Under such conditions, curve fitting of the “power law” expressions to the PSD functions estimates “slope” and “intercept” parameters of the seahorse click signal. Striking differences in the feeding click characteristics of various sizes of the H. kuda and between the sexes were discernable based on the clustering pattern among the estimated “slope” and “intercept” parameters. Further studies using the seahorse body sizes and weight (wet) with respect to the peak frequencies estimated from the PSD distribution show results equivalent to those from previous studies.