Acoustic Recognition Research Articles

Comparative performance analysis of three classifiers for acoustic signal-based recognition of motorcycles using time- and frequency-domain features

Vehicles of different types generate dissimilar sound patterns even in similar working conditions. In this study, the motorcycles are classified into bikes and scooters based on the sounds produced by them. Simple time-domain features and frequency-domain features are used for classifiers. The performances of artificial neural network, knowledge-based classifier and dynamic time warping are compared and reported. All these classifiers have shown more than 90% classification accuracy when trained with minimum 40% of the samples.

Acoustic Identification of Individuals within Large Avian Populations: A Case Study of the Brownish-Flanked Bush Warbler, South-Central China

Acoustic identification is increasingly being used as a non-invasive method for identifying individuals within avian populations. However, most previous studies have utilized small samples of individuals (<30). The feasibility of using acoustic identification of individuals in larger avian populations has never been seriously tested. In this paper, we assess the feasibility of using distinct acoustic signals to identify individuals in a large avian population (139 colour-banded individuals) of Brownish-flanked Bush Warbler (Cettia fortipes) in the Dongzhai National Nature Reserve, south-central China. Most spectro-temporal variables we measured show greater variation among individuals than within individual. Although there was slight decline in the correct rate of individual identification with increasing sample sizes, the total mean correct rate yielded by discriminant function analysis was satisfactory, with more than 98% of songs correctly recognized to the corresponding individuals. We also found that using a part of randomly selected measured variables was sufficient to obtain a high correct rate of individual identification. We believe that our work will increase confidence in the use of using acoustic recognition techniques for avian population monitoring programs.

A new acoustic target recognition method based on gaussian mixture model

Target recognition is one of the most important components of an acoustic detection system. Conventional acoustic target recognition algorithms base the classification results on features from a single frame of signal, suffering from inability to take the dynamic characteristics of acoustic signal into account, which can significantly increase the separability of acoustic targets. In this paper, we present a novel method for acoustic target recognition. Acoustic signal to be classified is divided into a sequence of signal segments. Gaussian mixture model is employed to model the sequence, and EM algorithm is utilized to learn the parameters. This idea is motivated by the interpretation that the Gaussian components can represent some general target-dependent spectral shapes. Experimental results show that the method achieves high recognition accuracy on different data sets, and is highly robust to ambient noise.

An acoustic system for the individual recognition of insects

Research into acoustic recognition systems for insects has focused on species identification rather than individual identification. In this paper, the feasibility of applying pattern recognition techniques to construct an acoustic system capable of automatic individual recognition for insects is investigated analytically and experimentally across two species of Orthoptera. Mel-frequency cepstral coefficients serve as the acoustic feature, and α-Gaussian mixture models were selected as the classification models. The performance of the proposed acoustic system is promising and displays high accuracy. The results suggest that the acoustic feature and classifier method developed here have potential for individual animal recognition and can be applied to other species of interest.

Acoustic targets feature extraction method based on manifold learning

A new method for feature extraction of acoustic targets based on manifold learning is proposed. The new method overcomes the shortcomings of the original manifold learning algorithm by minimising the impact of noise on the manifold and makes full use of the classification information to extract the intrinsic features of acoustic targets, thus improving the accuracy of acoustic targets recognition. Results of experiments suggest the proposed method is efficient and feasible.

Genome duplication in amphibians and fish: an extended synthesis

Whole genome duplication (leading to polyploidy) is widely accepted as an important evolutionary force in plants, but it is less recognized as a driver of animal diversification. Nevertheless, it occurs across a wide range of animals; this review investigates why it is particularly common in fish and amphibians, while rare among other vertebrates. We review the current geographic, ecological and phylogenetic distributions of sexually reproducing polyploid taxa before focusing more specifically on what factors drive polyploid formation and establishment. In summary, (1) polyploidy is phylogenetically restricted in both amphibians and fishes, although entire fish, but not amphibian, lineages are derived from polyploid ancestors. (2) Although mechanisms such as polyspermy are feasible, polyploid formation appears to occur principally through unreduced gamete formation, which can be experimentally induced by temperature or pressure shock in both groups. (3) External reproduction and fertilization in primarily temperate freshwater environments potentially exposes zygotes to temperature stress, which can promote increased production of unreduced gametes. (4) Large numbers of gametes and group breeding in relatively confined areas could increase the probability of compatible gamete combinations in both groups. (5) Both fish and amphibians have a propensity to form reproductively successful hybrids; although the relative frequency of autopolyploidy versus allopolyploidy is difficult to ascertain, multiple origins involving hybridization have been confirmed for a number of species in both groups. (6) Problems with establishment of polyploid lineages associated with minority cytotype exclusion could be overcome in amphibians via assortative mating by acoustic recognition of the same ploidy level, but less attention has been given to chemical or acoustic mechanisms that might operate in fish. (7) There is no strong evidence that polyploid fish or amphibians currently exist in more extreme environments than their diploid progenitors or have broader ecological ranges. (8) Although pathogens could play a role in the relative fitness of polyploid species, particularly given duplication of genes involved in immunity, this remains an understudied field in both fish and amphibians. (9) As in plants, many duplicate copies of genes are retained for long periods of time, indicative of selective maintenance of the duplicate copies, but we find no physiological or other reasons that could explain an advantage for allelic or genetic complexity. (10) Extant polyploid species do not appear to be more or less prone to extinction than related diploids in either group. We conclude that, while polyploid fish and amphibians share a number of attributes facilitating polyploidy, clear drivers of genome duplication do not emerge from the comparison. The lack of a clear association of sexually reproducing polyploids with range expansion, harsh environments, or risk of extinction could suggest that stronger correlations in plants may be driven by shifts in mating system more than ploidy. However, insufficient data currently exist to provide rigorous tests of these hypotheses and we make a plea for zoologists to also consider polyploidy as a possibility in continuing taxonomic surveys.

Acoustic Signal Recognition Based on Time Series Model

Using time series model, isometric transformation time series model and ARTAFIT model, we deal with acoustic signal, obtaining different sets of parameters according to different acoustic signals. We use support vector machine (SVM) to recognize different acoustic signals by analyzing different sets of parameters. When the parameter set is too large, we should first reduce order making use of principal component analysis (PCA), then we can recognize them using support vector machine. In the end, we give a case study, which indicate the results of applying our models are satisfactory.

A novel approach to research on feature extraction of acoustictargets based on manifold learning

In order to overcome the deficiency of robustness of low altitude passive acoustic target recognition, the manifold learning is applied to the feature extraction of acoustic targets. Based on the classical algorithm of manifold learning, in the paper we study and discuss the low-dimensional manifold in the frequency-domain of acoustic signals. This method is used to solve the target recognition problem with two data sets to verify its effectiveness, after which the performance is analyzed. The result indicates that the manifold learning can discover the intrinsic feature and increase the accuracy and the robustness of low altitude passive acoustic target recognition system.

HOW AND WHEN DO LAMBS RECOGNIZE THE BLEATS OF THEIR MOTHERS?

ABSTRACT In domestic sheep Ovis aries, the mother and the young display a preferential bond for each other that relies on multimodal inter-individual recognition. Lambs show a preference for their own dam shortly after birth, and this is important for their survival. The role of acoustic cues in this early preference for the mother is not clear. The aim of the present work was to analyze the timing of acoustic recognition of the mother and to identify the physical parameters used in the recognition of maternal bleats by the lamb. In a first study, we investigated the ability of lambs to discriminate between the bleats of their own mother and an alien equivalent mother in a two-choice test. Both ewes were hidden behind a canvas sheet and lambs were not allowed to approach the dams closer than 1 m, thus preventing visual as well as olfactory perception. Tests were conducted 12 hr, 24 hr or 48 hr after birth (n=19 or 20/group). An indication of vocal discrimination was already present at 24 hr and at 48 hr lambs spent significantly more time near their mother than near the alien dam. In a second step, we investigated which physical parameters of the bleats were important for recognition. For this, we conducted playback experiments with modified bleats at two weeks postpartum. Ours results show that lambs pay attention to a combination of various time, energy and frequency parameters: timbre (distribution of energy within the spectrum), amplitude and frequency modulations appear to be the most important parameters encoding the individual signature. We conclude that vocal recognition between the ewe and her lamb plays an important role in the display of preferential mother-young bond from very early on. Our studies also demonstrate that the encoding of the individual signature is not limited to the frequency domain but rather involves a multiparametric encoding process.

Early warning of hazard for pipelines by acoustic recognition using principal component analysis and one-class support vector machines

This paper proposes a method for early warning of hazard for pipelines. Many pipelines transport dangerous contents so that any damage incurred might lead to catastrophic consequences. However, most of these damages are usually a result of surrounding third-party activities, mainly the constructions. In order to prevent accidents and disasters, detection of potential hazards from third-party activities is indispensable. This paper focuses on recognizing the running of construction machines because they indicate the activity of the constructions. Acoustic information is applied for the recognition and a novel pipeline monitoring approach is proposed. Principal Component Analysis (PCA) is applied. The obtained Eigenvalues are regarded as the special signature and thus used for building feature vectors. One-class Support Vector Machine (SVM) is used for the classifier. The denoising ability of PCA can make it robust to noise interference, while the powerful classifying ability of SVM can provide good recognition results. Some related issues such as standardization are also studied and discussed. On-site experiments are conducted and results prove the effectiveness of the proposed early warning method. Thus the possible hazards can be prevented and the integrity of pipelines can be ensured.

A call-independent and automatic acoustic system for the individual recognition of animals: A novel model using four passerines

Research into acoustic recognition systems for animals has focused on call-dependent and species identification rather than call-independent and individual identification. Here we present a system for automatic call-independent individual recognition using mel-frequency cepstral coefficients and Gaussian mixture models across four passerine species. To our knowledge this is the first application of these techniques to the individual recognition of birds, and the results are promising. Accuracies of 89.1–92.5% were achieved and the acoustic feature and classifier method developed here have excellent potential for individual animal recognition and can be easily applied to other species.

Special Issue on Advances on Intelligent Multimedia Processing

As computer and sensor technology advances make increasing amounts of multimedia data available, multimedia processing methodologies are needed for multimodal data fusion, efficient data processing, information extraction, and added-value data generation. This special issue introduces the following latest practical developments in image processing, acoustical signal processing, pattern recognition, data mining, and visualization: S. Kobashi et al. propose a robust algorithm for reconstructing total knee arthroplasty implants from X-ray cone-beam images. Y. Hatakeyama et al. propose an algorithm for classifying ultrasonic abdominal images with the help of reports in text format. T. Miyazaki et al. analyze breathy and rough speech by the elderly. M. Nii et al. present a genetic algorithm for classifying nursing-care text. These papers provide useful insights into medical diagnosis and nursing care in the aging society coming to dominate 21st century. T. Toyota et al. visualize Japanese law networks based on granular computing. K. Sawase et al. present a management system for large databases of tagged images. These graphical user interface techniques will be helpful to those who are not computer experts. H. Kawano et al. propose a classification algorithm for segmenting range data into multiple quadric surfaces. Y. Arai et al. propose a nearest-neighbor method for personal authentication. H. Kawano et al. present an algorithm for extracting the structure of decorative characters based on graph spectral decomposition. These classification and discriminant algorithms provide a basis for multimedia data processing. G. Tanaka et al. propose a color transfer algorithm. H. Orii et al. present an image completion algorithm. G. Tanaka et al. present an image enhancement algorithm to noisy images. M. Mizumachi et al. propose a stochastic algorithm for estimating sound source direction. These image and acoustical processing algorithms improve the quality of digital data and will provide new applications in these areas. We thank the authors and referees whose invaluable work and kind comments have made this special issue possible and improved overall paper quality.

Early vocal recognition of mother by lambs: contribution of low- and high-frequency vocalizations

In sheep, Ovis aries, lambs show a preference for their own mother soon after birth, which is important for their survival. The relative role of low- and high-pitched bleats provided by the mother in this early preference is not clear. While acoustic recognition is clearly established in 2-day-old lambs, it is not known whether lambs use an individual acoustic signature, or some other feature of acoustic communication (type of bleat or a combination of responses) by the mother indicating acceptance. The acoustic analysis of bleats demonstrated that both types of vocalization can support individual discrimination and that frequency spectrum and amplitude modulation exhibit an individual stereotypy. However, the characteristics of bleats and the quantity of encoding information in them differed between low- and high-pitched bleats. In two-choice procedures, we tested the ability of 48 h-old lambs to discriminate between recorded bleats of their own mother and an unfamiliar equivalent mother. Lambs showed a preference for the playback of low-pitched bleats of their own mother, but not for high-pitched bleats. Therefore, 48 h-old lambs can use an individual acoustic signature to recognize their own mother when low-pitched bleats are used. In contrast when high-pitched bleats are used, a vocal exchange between the mother and her lamb appears necessary. We conclude that 48 h-old lambs are able to learn and recognize the individual characteristics of their mother's low-pitched bleats and that this vocal recognition may be important for the establishment of the mother–young bond even at this early stage of the relationship.

An investigation in acoustic insect recognition

This study aims to provide general technicians who manage pests in production with a convenient way to recognize insects. Several viable schemes to identify insect sounds automatically are introduced using sound parameter standardization techniques that dominate speaker recognition technology. The acoustic signal is preprocessed, segmented into a series of sound samples. Melfrequency cepstrum coefficient (MFCC) and Sub-band based cepstral (SBC) are extracted, respectively from the sound samples, and Vector Quantization(VQ) codebook and Hidden Markov Model(HMM) are trained with given features. The matching for a test sample is completed by finding the nearest neighbour in all the VQ codebooks or the best matcher in all HMMs. These methods are evaluated and compared in a database with acoustic samples of 70 different insect sounds. The MFCC and HMM based methods demonstrated their better performance, whose recognition accuracy exceeds 98%.

Rub-impact Acoustic Emission Signal Recognition of Rotating Machinery Based on Gaussian Mixture Model

On the basis of modal acoustic emission and narrow-band signal theory, the mathematical expression of multi-modal AE signal is given. A mixed parameter composed of logarithm cepstral coefficients and fractal dimension is presented as the characteristic coefficients of rub-impact AE signal, and the AE recognition system based on Gaussian mixture model is also established. Because the eigenvector of each modal wave in rub-impact AE signal is different, so a model of probability density function of eigenvector is built and it is regarded as a clustering. Training model is established with mean, covariance matrix and probability of each clustering. Sum likelihood probability weighted by maximal ratio combining to likelihood probability of Gaussian model will be obtained when the tested signal is recognized, if the probability is larger than a set threshold, it can be confirmed that rub-impact AE is available. Rub impact AE signals are sampled from rotor test stand, the types of input model of Gaussian are sorted by AE wavrform shapes and its fractal dimensions, then white noise and non-stationary noise are added to simulate the real AE signal. In the end, this Gaussian mixture model is used for AE recognition. The results indicate that the model has high recognition rate and good anti-noise ability.

Response to displaced neighbours in a territorial songbird with a large repertoire

Neighbour recognition allows territory owners to modulate their territorial response according to the threat posed by each neighbour and thus to reduce the costs associated with territorial defence. Individual acoustic recognition of neighbours has been shown in numerous bird species, but few of them had a large repertoire. Here, we tested individual vocal recognition in a songbird with a large repertoire, the skylark Alauda arvensis. We first examined the physical basis for recognition in the song, and we then experimentally tested recognition by playing back songs of adjacent neighbours and strangers. Males showed a lower territorial response to adjacent neighbours than to strangers when we broadcast songs from the shared boundary. However, when we broadcast songs from the opposite boundary, males showed a similar response to neighbours and strangers, indicating a spatial categorisation of adjacent neighbours' songs. Acoustic analyses revealed that males could potentially use the syntactical arrangement of syllables in sequences to identify the songs of their neighbours. Neighbour interactions in skylarks are thus subtle relationships that can be modulated according to the spatial position of each neighbour.

Bite Weight Prediction From Acoustic Recognition of Chewing

Automatic dietary monitoring (ADM) offers new perspectives to reduce the self-reporting burden for participants in diet coaching programs. This paper presents an approach to predict weight of individual bites taken. We utilize a pattern recognition procedure to spot chewing cycles and food type in continuous data from an ear-pad chewing sound sensor. The recognized information is used to predict bite weight. We present our recognition procedure and demonstrate its operation on a set of three selected foods of different bite weights. Our evaluation is based on chewing sensor data of eight healthy study participants performing 504 habitual bites in total. The sound-based chewing recognition achieved recalls of 80% at 60%-70% precision. Food classification of chewing sequences resulted in an average accuracy of 94%. In total, 50 variables were derived from the chewing microstructure, and were analyzed for correlations between chewing behavior and bite weight. A subset of four variables was selected to predict bite weight using linear food-specific models. Mean weight prediction error was lowest for apples (19.4%) and largest for lettuce (31%) using the sound-based recognition. We conclude that bite weight prediction using acoustic chewing recordings is a feasible approach for solid foods, and should be further investigated.

Innate or learned acoustic recognition of avian predators in rodents?

Calls of avian predators potentially predict danger for murine rodents. Indeed, exposure of field-experienced rodents to owl calls induces defensive behavior suggesting that recognition of vocalizations of avian predators is innate. To address this hypothesis, we investigated whether laboratory-reared and predator-naive rodents (mice, gerbils, rats) express defensive behavior in response to calls of different avian predators but we observed no such defensive behavior. We then asked whether the calls of avian predators are faster or better learned as a danger-predicting cue than the calls of avian non-predators. All calls could be learned as danger-predicting cues, but we found no differences in the speed or strength of the learning. Taken together, our results suggest that there is no innate recognition of the calls of avian predators in murine rodents and that the recognition of the calls of avian predators observed in field-experienced rodents is acquired by learning.

Mother–young vocal communication and acoustic recognition promote preferential nursing in sheep

In mammals with precocial neonates, exclusive maternal care and investment depend on mutual mother-young recognition. In sheep, this is ensured by rapid olfactory recognition of the neonate by its mother. However, recent studies suggest that other processes may participate in preferential maternal care. We investigated the possibility that acoustic communication promotes preferential nursing of the lamb. In the first of two studies, we examined the association between nursing and vocal activity in ewes and their lambs during the first 2 weeks of lactation. As early as 3 and 6 h postpartum, vocal activity was preferentially concentrated before nursing and by day 15 all vocal activity occurred within 2.5 min before nursing. In the second study, we tested the responses of ewes and lambs at 15 days postpartum to the playback of recorded bleats of their partner or from unrelated ewes and lambs. When playback was performed 30 min after a nursing episode, both ewes and lambs responded to bleats of their partner but not to bleats from alien subjects. When playback was performed 5 min after nursing, ewes did not respond to any lamb's bleats, while lambs continued responding to the bleats of their mothers, but significantly less than 30 min after nursing. Nursing therefore appears to play an important role in structuring very early vocal communication between the mother and her neonate. In turn, if the motivational state of the members of the mother-young dyad is adequate, this ensures the display of mutual acoustic recognition and prepares them for preferential nursing before maternal olfactory recognition of the lamb comes into play.

Recognition of potential danger to buried pipelines based on sounds

In order to prevent damages to the underground pipelines, automatic, continuous, and low-cost monitoring systems are becoming indispensable. Recent reports showed that most pipeline damage is caused by the third-party activities, which can be identified by detecting for dangerous construction equipments nearby, rather than the material failure and corrosion. This paper focuses on acoustically recognizing road cutters since they prelude most construction activities in modern cities. In this paper, a novel monitoring mechanism is proposed for sound-based pipeline monitoring system. Instead of the conventional approach of using overlapped frames, this paper proposes segmenting sound into separate frames, with an interval between every two frames. After gathering sound samples, a tag stack decision maker is also proposed to make the final decision. Different features of two acoustic recognition technologies, Mel frequency cepstral coefficients (MFCCs) and linear prediction coding cepstrum (LPCC), were studied, tried, and compared. For the classifiers, both the Euclidean cepstral distance and one-class support vector machine (SVM) were studied. Real site experiments were conducted and the resulting data were analyzed. Recognition success rate at different conditions were studied and compared. Our results showed that MFCC feature performs the best. Cepstral Euclidean distance is simple but effective, while one-class SVM can help to extend the classification capabilities. The sound recognition technologies studied in this paper will be very useful for future pipeline monitoring systems to prevent accidental breakage, thus ensuring the safety of underground lifeline infrastructures. Copyright © 2008 John Wiley & Sons, Ltd.