Acoustic Units Research Articles

Holocene stratigraphy of the Ångermanälven River estuary, Bothnian Sea

This study explores the Holocene depositional succession at the IODP Expedition 347 sites M0061 and M0062 in the vicinity of the Angermanalven River estuary in the Bothnian Sea sector of the Baltic Sea in northern Scandinavia. Site M0061 is located in a coastal offshore setting (87.9 m water depth), whereas site M0062 is fully estuarine (69.3 m water depth). The dataset comprises acoustic profiles and sediment cores collected in 2007 and late 2013 respectively. Three acoustic units (AUs) were recognized. Lowermost AU1 is interpreted as a poorly to discontinuous stratified glaciofluvial deposit, AU2 as a stratified conformable drape of glaciolacustrine origin, and AU3 as a poorly stratified to stratified mud drift. A strong truncating reflector separates AU2 and AU3. Three lithological units (LUs) were defined in the sediment cores. LU1 consists of glaciofluvial sand and silt gradating into LU2, which consists of glaciolacustrine varves. A sharp contact interpreted as a major unconformity separates LU2 from the overlying LU3 (brackish-water mud). In the basal part of LU3, one debrite (site M0061) or two debrites (site M0062) were recognized. Information yielded from sediment physical properties (magnetic susceptibility, natural gamma ray, dry bulk density), geochemistry (total carbon, total organic carbon, total inorganic carbon and nitrogen), and grain size support the LU division. The depositional succession was formally subdivided into two alloformations: the Utansjo Alloformation and overlying Hemson Alloformation; the Utansjo Alloformation was further subdivided into two lithostratigraphic formations: the Storfjarden and Abordson formations. The Storfjarden (sandy outwash) and Abordson (glaciolacustrine rhythmite) formations represent a glacial retreat systems tract, which started at ca. 10.6 kyr BP. Their deposition was mainly controlled by meltwater from the retreating ice margin, glacio-isostatic land uplift and the regressive (glacial) lake level. The Hemson Alloformation (organic-rich brackish-water mud) represents a period of forced regression, starting possibly at ca. 9.5 kyr BP. At about 7 kyr BP, brackish water reached the study area as a result of the mid-Holocene marine flooding of the Baltic Sea Basin, but the rapid land uplift soon surpassed the associated (Littorina) transgression. Changed near-bottom current patterns, caused by the establishment of a permanent halocline, and the reduced sediment consistency caused by increased organic deposition resulted in a sharp and erosional base of the brackish-water mud. Estuarine processes and salinity stratification at site M0062 started to play a more important role. This study applies a combined allostratigraphic and lithostratigraphic approach over the conventional Baltic Sea stages. This approach makes it more straightforward to study this Baltic Sea deglaciation–postglacial sequence and compare it to other formerly glaciated shallow sea estuaries.

Acoustically diverse vocalization repertoire in the Himalayan leaf-nosed bat, a widely distributed Hipposideros species.

Insectivorous bats vocalize to both communicate with conspecifics and to echolocate. The communicative vocalizations or "calls" of bats either consist of or are constructed from discrete acoustic units, termed "syllables." This study examined syllable diversity in the Himalayan leaf-nosed bat, Hipposideros armiger, a species that is widely distributed across Southeast Asia. This social species' vocalizations were hypothesized to consist of a wide variety of syllables facilitating its social interactions. To test this hypothesis, multiple acoustic parameters were measured from recorded vocalizations to map the acoustic boundaries of syllables. Spectrographic signatures were used to classify all recorded sounds into 35 distinct syllable types-18 as simple syllables and 17 as composites. K-means clustering independently provided an optimal fit of simple syllables into 18 clusters with a good correspondence to 15 spectrographically assigned syllable types. Discriminant analysis further confirmed the spectrographic classification of constant frequency syllables (0% misclassification) and revealed a low (<15%) misclassification of spectrograms for all examples of frequency modulation syllables. Multidimensional scaling of mean values of multiple parameters provided a spectrographically constrained relational mapping of syllable types within two dimensions. These data suggest that H. armiger has a complex, well organized syllabic repertoire despite simple syllables being rarely emitted in isolation.

Buried iceberg-keel scouring on the southern Spitsbergenbanken, NW Barents Sea

PARASOUND (3.5kHz) subbottom echosounder profiles acquired on the southern Spitsbergenbanken, NW Barents Sea, show iceberg-keel scouring features which are buried by sediment that accumulated during the post Last Glacial Maximum (LGM) sea-level rise. Four acoustic units (Units 1 to 4 in stratigraphic order) were differentiated, based on the characterization of their acoustic facies and reflection surfaces. Unit 1 shows a chaotic internal structure and is interpreted as a glacial till, whereas the laminated Units 2 to 4 accumulated by sediment settling from suspension clouds and bottom currents during the last deglaciation phase. The top of Unit 2 was frequently incised by iceberg keels, resulting in up to 12m deep ploughmarks which were later filled and buried by Unit 3 and 4 sediments. Three main paleo-evironmental changes controlled the evolution of the facies succession: (1) The major shift from till formation (Unit 1) below grounded ice to the accumulation of laminated sediments (Unit 2) which are inferred to reflect ice lifting and meltwater release; (2) Iceberg-keel scouring after sedimentation of Unit 2; (3) the probable abrupt termination of iceberg-keel scouring related to the glacio-eustatic sea-level rise. A linkage between these episodes of changes and short-lasting phases of rapid post LGM sea-level rise, such as meltwater pulses, is inferred, although further studies are needed to better understand the temporal and genetic relationships between the sedimentary events recognized in the Barents Sea and climate changes.

Call concatenation in non-primate mammals

Animals produce a variety of call sequences, from simple repetitions of the same acoustic unit to the combination of different, meaningful acoustic units resulting in a new meaning. On the example of several highly vocal mongoose species, including meerkats (Suricata suricatta), and banded mongoose (Mungos mungo) I identify the composition of different acoustic combinations and relate them to their function and meaning. I discuss this in regards to the underlying cognitive mechanisms and the differences and similarities in relation to other combinatorial communication systems in birds, but particularly shown for primates, and also to human language.

A helical-structured acoustic metamaterial

Here, we would like to present a helical-structured acoustic metamaterial that can effectively decelerate sound propagation. Unlike currently avaliable solutions which are mostly dispersive due to the requirement of local resonators., this non-dispersive helical-structured metamaterials is able to provide high effective refractive index. By adjusting the helicity of the acoustic metamaterial structures, we can further modify such refractive index. The performance of this helical-structured acoustic metamaterial has been numerically and experimentally verified. We even designed a lens base on multiple inhomogeneous helical-structured acoustic metamaterial unit cells. This lens can turn normally travelling plane wave into a self-accelerating beam.

On the Use of Acoustic Unit Discovery for Language Recognition

In this paper, we explore the use of large-scale acoustic unit discovery for language recognition. The deep neural network-based approaches that have achieved recent success in this task require transcribed speech and pronunciation dictionaries, which may be limited in availability and expensive to obtain. We aim to replace the need for such supervision via the unsupervised discovery of acoustic units. In this work, we present a parallelized version of a Bayesian nonparametric model from previous work and use it to learn acoustic units from a few hundred hours of multilingual data. These unit (or senone) sequences are then used as targets to train a deep neural network-based i-vector language recognition system. We find that a score-level fusion of our unsupervised system with an acoustic baseline can shrink the gap significantly between the baseline and a supervised benchmark system built using transcribed English. Subsequent experiments also show that an improved acoustic representation of the data can yield substantial performance gains and that language specificity is important for discovering meaningful acoustic units. We validate the generalizability of our proposed approach by presenting state-of-the-art results that exhibit similar trends on the NIST Language Recognition Evaluations from 2011 and 2015.

Preparation and Properties of Foamed Cellulose-Polymer Microsphere Hybrid Materials for Sound Absorption

Sustainability and eco-efficiency are presently directing the development of the next generation of acoustic materials. In this work, foamed cellulose-polymer microsphere (PM) hybrid materials, having sound-absorbing capability, were prepared by incorporating the PMs into cellulose fibers by dehydration and foaming processes. The evolution in morphology of PMs during foaming process was investigated for different heating temperatures. The beating process disintegrated the microscopic cellulose fiber into the smaller fibers, which connected the PMs by a unique fibrous network. The influences of foaming temperature, PM content, and total areal density on the sound absorbing property of composites were studied. The results showed that incorporating the acoustic unit of elastic PMs into the porous cellulose fiber-based network significantly improved the sound absorbing ability of the composites. The sound-absorbing hybrid materials appear to be a promising alternative to non-degradable organic or inorganic acoustic composites, being economical, simple, and eco-friendly.

Nocturnal songs in a diurnal passerine: attracting mates or repelling intruders?

We addressed the role of nocturnal singing in the diurnal field sparrow, Spizella pusilla. We were particularly interested in whether nocturnal song is used to repel intruders (intrasexual function) and/or to attract females seeking extrapair copulations (intersexual function). First, we used autonomous acoustic recording units (ARUs) and an automated detection and classification system to examine the associations of singing behaviour of mated field sparrows with social factors (fertility stage, presence of neighbour song and presence of intruder song). Second, using an automated radiotelemetry system (ARTS), autonomous acoustic recording units (ARUs) and automated playback systems, we conducted a nocturnal playback experiment to explore how mated male and female field sparrows respond to nocturnal songs at night and across prefertile, fertile and postfertile stages. We found that neighbours and intruder males tended to sing complex songs on the same nights when focal males sang. While not all focal males sang at night, all male birds (resident males, neighbours and intruders) that sang tended to sing according to the fertility periods of the resident male's female. Our playback experiment demonstrated that, although field sparrows usually slept and were inactive at night, they moved in response to nocturnal field sparrow songs more than to control stimuli (other species vocalizations). Interestingly, we found that males did not respond by singing or countersinging to field sparrow nocturnal songs (simulated intrusions). We also demonstrated that male activity responses were similar during the prefertile and postfertile stages, while females responded more during the prefertile and fertile stages than during the postfertile stage. Consequently, the nocturnal song in the field sparrow appears to play a role in extrapair mate attraction (intersexual function) more than in repelling intruders (intrasexual function). While nocturnal singing is an infrequent behaviour of most diurnal birds, it may be important in the mating system of these birds.

Seasonal and daily patterns of nocturnal singing in the Field Sparrow (Spizella pusilla)

Nocturnal singing in diurnal birds is poorly described in the literature and is not well understood. Most research has been limited to the documentation of the behavior, with little information regarding the patterns and function of nocturnal song. We used autonomous acoustic recording units and automated detection and classification algorithms to examine the seasonal and nocturnal patterns of song in the Field Sparrow, Spizella pusilla, a diurnal songbird. We used a generalized linear mixed model to investigate the patterns in the probability of recording each of the two song types described for the species (simple and complex) at night from 7938 10-min recordings from 14 different individuals singing in six different grassland patches. We fit different models investigating the influence of date within the season and time period during the night, and compared models using the Bayesian information criterion. Male Field Sparrows sang both simple and complex songs during the early stages of territory settlement and mate acquisition. However, as the breeding season progressed, we detected fewer instances of simple song, while the occurrence of complex song increased, reaching its peak during the height of reproductive activity. The seasonal pattern of simple and complex songs suggests that they may have a similar function at night and during the day, with simple songs used for inter-sexual interactions and complex songs used for intrasexual functions. Consequently, the role and function of nocturnal song may be more important for reproductive activities than previously assumed.

Automatic non-destructive three-dimensional acoustic coring system for in situ detection of aquatic plant root under the water bottom

Digging is necessary to detect plant roots under the water bottom. However, such detection is affected by the transparency of water and the working skills of divers, usually requires considerable time for high-resolution sampling, and always damages the survey site. We developed a new automatic non-destructive acoustic measurement system that visualizes the space under the water bottom, and tested the system in the in situ detection of natural plant roots. The system mainly comprises a two-dimensional waterproof stage controlling unit and acoustic measurement unit. The stage unit was electrically controlled through a notebook personal computer, and the space under the water bottom was scanned in a two-dimensional plane with the stage unit moving in steps of 0.01 m (±0.0001 m). We confirmed a natural plant root with diameter of 0.025–0.030 m in the reconstructed three-dimensional acoustic image. The plant root was at a depth of about 0.54 m and the propagation speed of the wave between the bottom surface and plant root was estimated to be 1574 m/s. This measurement system for plant root detection will be useful for the non-destructive assessment of the status of the space under the water bottom.

A Permanent Automated Real-Time Passive Acoustic Monitoring System for Bottlenose Dolphin Conservation in the Mediterranean Sea.

Within the framework of the EU Life+ project named LIFE09 NAT/IT/000190 ARION, a permanent automated real-time passive acoustic monitoring system for the improvement of the conservation status of the transient and resident population of bottlenose dolphin (Tursiops truncatus) has been implemented and installed in the Portofino Marine Protected Area (MPA), Ligurian Sea. The system is able to detect the simultaneous presence of dolphins and boats in the area and to give their position in real time. This information is used to prevent collisions by diffusing warning messages to all the categories involved (tourists, professional fishermen and so on). The system consists of two gps-synchronized acoustic units, based on a particular type of marine buoy (elastic beacon), deployed about 1 km off the Portofino headland. Each one is equipped with a four-hydrophone array and an onboard acquisition system which can record the typical social communication whistles emitted by the dolphins and the sound emitted by boat engines. Signals are pre-filtered, digitized and then broadcast to the ground station via wi-fi. The raw data are elaborated to get the direction of the acoustic target to each unit, and hence the position of dolphins and boats in real time by triangulation.

High-Resolution Analysis of Seismic Air Gun Impulses and Their Reverberant Field as Contributors to an Acoustic Environment.

In September and October 2011, a seismic survey took place in Baffin Bay, Western Greenland, in close proximity to a marine protected area (MPA). As part of the mitigation effort, five bottom-mounted marine acoustic recording units (MARUs) collected data that were used for the purpose of measuring temporal and spectral features from each impulsive event, providing a high-resolution record of seismic reverberation persistent after the direct impulse. Results were compared with ambient-noise levels as computed after the seismic survey to evidence that as a consequence of a series of repeating seismic impulses, sustained elevated levels create the potential for masking.

Leak detection and localization using acoustic emission technique

Article history: Received 6 April, 2015 Accepted 10 October 2015 Available online 11 October 2015 This article provides a method of acoustic emission (AE) technique to detect leakage in pipeline and locate the position of the leakage. The AE sensor is made of piezoelectric effect transducer to pick up the acoustic emission signal which is generated from the turbulent flow at the leak position. The signal conditioning unit is used to enhanced and eliminate the background noise from the leak location sources. The main acoustic emission processing unit is used to acquire and process the extracted AE characteristic parameters from preprocessing waveform. The leak pipeline is simulated by drilling the hole and plugged with M8 screw at different locations. The results show that the proposed AE method can detect and locate simultaneous leak condition in pipeline with promising results. © 2016 Growing Science Ltd. All rights reserved.

Unsupervised Linear Discriminant Analysis for Supporting DPGMM Clustering in the Zero Resource Scenario

In this work we make use of unsupervised linear discriminant analysis (LDA) to support acoustic unit discovery in a zero resource scenario. The idea is to automatically find a mapping of feature vectors into a subspace that is more suitable for Dirichlet process Gaussian mixture model (DPGMM) based clustering, without the need of supervision. Supervised acoustic modeling typically makes use of feature transformations such as LDA to minimize intra-class discriminability, to maximize inter-class discriminability and to extract relevant informations from high-dimensional features spanning larger contexts. The need of class labels makes it difficult to use this technique in a zero resource setting where the classes and even their amount are unknown. To overcome this issue we use a first iteration of DPGMM clustering on standard features to generate labels for the data, that serve as basis for learning a proper transformation. A second clustering operates on the transformed features. The application of unsupervised LDA demonstrably leads to better clustering results given the unsupervised data. We show that the improved input features consistently outperform our baseline input features.

Variational Inference for Acoustic Unit Discovery

Recently, several nonparametric Bayesian models have been proposed to automatically discover acoustic units in unlabeled data. Most of them are trained using various versions of the Gibbs Sampling (GS) method. In this work, we consider Variational Bayes (VB) as alternative inference process. Even though VB yields an approximate solution of the posterior distribution it can be easily parallelized which makes it more suitable for large database. Results show that, notwithstanding VB inference is an order of magnitude faster, it outperforms GS in terms of accuracy.

Study of Large Data Resources for Multilingual Training and System Porting

This study investigates the behavior of a feature extraction neural network model trained on a large amount of single language data (“source language”) on a set of under-resourced target languages. The coverage of the source language acoustic space was changed in two ways: (1) by changing the amount of training data and (2) by altering the level of detail of acoustic units (by changing the triphone clustering). We observe the effect of these changes on the performance on target language in two scenarios: (1) the source-language NNs were used directly, (2) NNs were first ported to target language.The results show that increasing coverage as well as level of detail on the source language improves the target language system performance in both scenarios. For the first one, both source language characteristic have about the same effect. For the second scenario, the amount of data in source language is more important than the level of detail.The possibility to include large data into multilingual training set was also investigated. Our experiments point out possible risk of over-weighting the NNs towards the source language with large data. This degrades the performance on part of the target languages, compared to the setting where the amounts of data per language are balanced.

Improvement of methods for estimation of catchability for trawl survey systems with use of hydroacoustic means (on example of walleye pollock in the Okhotsk Sea)

In commercial fishery, catching properties of fishing gears are developed mostly for effective catch of fish, rather than for precise measurement of fish distribution density in water, so a problem of catchability exists and this parameter should be determined for each fishing gear. Various methods for estimation of catchability for trawl survey systems are considered. The experimental method of its estimation (comparing of fish density in towed volume determined by other tools with trawl catch) is considered as the most authentic. Scheme and algorithm of the measurements and calculations are described for standard hydroacoustic tools, as scientific echosounder ЕК-60, and software is developed for recalculation of trawl catch to equivalent acoustic units of fish density in the towed volume of water. Results of the catchability estimation for the trawl RT/TM 57/360 used for walleye pollock stock assessment in the Okhotsk Sea are analyzed. Mean value of the catchability coefficient in this case varies from 0.42 to 0.81, generally increasing with depth but decreasing both at the sea surface and on the depth &gt; 200 m. These results correspond well with the model of fish behavior in the noise field of vessel: the low catchability at the sea surface and its increasing with depth could be caused by avoiding reaction of fish to the noise and its moving to deeper layers. Besides, the catchability coefficient for pollock depends on fish size and time of day: it increases with the fish length but decreases for the largest individuals. Influence of the noise field of vessel on selectivity of survey trawls is modeled and there is shown that the depths of the maximum concentration of small-sized and large-sized pollock do not coincide. The developed software and received results are universal means for estimation of the catchability for trawls in all conditions of trawling, they allow to choose optimum mode for survey trawling in dependence on fish species, depth of trawling, and type of vessels and trawl.

Integrating human emotions with spatial speech using optimized selection of acoustic phonetic units

Abstract Synthesis of natural sounding speech is state of the art in the field of speech technology. Imitation of the dynamic human voice is required to generate this. The aim of this work is to develop and deploy a natural speech synthesizer for visually impaired persons. The synthesizer has been developed via an integrated approach of adding localization in expressive speech using a personalized speech corpus. A genetic algorithm has been implemented for optimal selection of acoustic phonetic units of speech. This concept has many applications. We tested one of those applications here in different aspects. Its performance is compared on various categories of listeners using a subjective listening test. Encouraging results on various parameters are received from visually impaired listeners.

Long recording sequences: how to track the intra-individual variability of acoustic signals.

Recently developed acoustic technologies - like automatic recording units - allow the recording of long sequences in natural environments. These devices are used for biodiversity survey but they could also help researchers to estimate global signal variability at various (individual, population, species) scales. While sexually-selected signals are expected to show a low intra-individual variability at relatively short time scale, this variability has never been estimated so far. Yet, measuring signal variability in controlled conditions should prove useful to understand sexual selection processes and should help design acoustic sampling schedules and to analyse long call recordings. We here use the overall call production of 36 male treefrogs (Hyla arborea) during one night to evaluate within-individual variability in call dominant frequency and to test the efficiency of different sampling methods at capturing such variability. Our results confirm that using low number of calls underestimates call dominant frequency variation of about 35% in the tree frog and suggest that the assessment of this variability is better by using 2 or 3 short and well-distributed records than by using samples made of consecutive calls. Hence, 3 well-distributed 2-minutes records (beginning, middle and end of the calling period) are sufficient to capture on average all the nightly variability, whereas a sample of 10 000 consecutive calls captures only 86% of it. From a biological point of view, the call dominant frequency variability observed in H. arborea (116Hz on average but up to 470 Hz of variability during the course of the night for one male) challenge about its reliability in mate quality assessment. Automatic acoustic recording units will provide long call sequences in the near future and it will be then possible to confirm such results on large samples recorded in more complex field conditions.

A comparative study on selecting acoustic modeling units in deep neural networks based large vocabulary Chinese speech recognition

This paper compared the performance of different acoustic modeling units in deep neural networks (DNNs) based large vocabulary continuous speech recognition (LVCSR) systems for Chinese. Recently, the deep neural networks based acoustic modeling method has achieved very competitive performance for many speech recognition tasks, and has become the focus of current LVCSR research. Some previous work have studied the context independent and context dependent DNNs based acoustic models. For Chinese, a syllabic language, the choice of basic modeling units under the background of DNNs based LVCSR systems is a very important issue.Three basic modeling units, syllables, initial/finals, phones, are discussed and compared. Experimental results show that, in the DNNs based systems, the context dependent (CD) phones obtain the best performance, and the context independent (CI) syllables have the similar performance with the CD initial/finals. How the number of clustered states impacts on the performance of DNNs based systems is also discussed, which showed different properties from the GMMs based systems. Besides, through introducing the multi-task learning strategy, these multiple modeling units can be combined in the DNNs training procedure. The experimental results indicate that combining these multiple modeling units using multi-task learning outperforms each individual modeling unit.