Classification Of Acoustic Signals Research Articles

Classification of acoustic signals using the statistics of the 1-D wavelet transform coefficients

The representation of an acoustic signal in terms of the wavelet sub-band coefficients is studied using their statistical features. This type of representation is suggested as an alternative tool for tomographic or geoacoustic inversions. Using a set of representative shallow water environments, the variation of the statistical behavior of the sub-band coefficients is associated with corresponding variations of the environmental parameters. Low frequency signals suitable for ocean acoustic tomography are simulated. It is shown that the statistics of the sub-band coefficients are best described using non-Gaussian heavy-tailed distributions such as those of the alpha-stable family. The variations of the distribution parameters are measured using special statistical similarity functions such as the Kullback-Leibler divergence which has been adopted in the present study. The distances between the statistical parameters of a given and reference signals determined by these functions, can be used for subsequent non-linear inversions aiming at the recovery of the environmental parameters.

IDENTIFICATION OF ACOUSTIC SIGNATURES FOR VEHICLES VIA REDUCTION OF DIMENSIONALITY

In this paper we propose a robust algorithm that solves two related problems: (1) Classification of acoustic signals emitted by different moving vehicles. The recorded signals have to be identified to which pre-existing group they belong to independently of the recording surrounding conditions. (2) Detection of the presence of a vehicle in a certain class via analysis of its acoustic signature against the existing database of recorded and processed acoustic signals. To achieve this detection with minimal false alarms we construct the acoustic signature of a certain vehicle using the distribution of the energies among blocks which consist of coefficients of multiscale local cosine transform (LCT) applied in the frequency domain of the acoustic signal. The proposed algorithm is robust even under severe noise and diverse rough surrounding conditions. This is a generic technology, which has many algorithmic variations, can be used to solve wide range of classification and detection problems which are based on a unique derivation of signatures.

Recognition of information-bearing elements in speech

An acoustic speech signal carries many different kinds of information: the basic linguistic message, many characteristics of the speaker of the message, details of the environment in which the message was produced and transmitted, etc. The human auditory/cognitive system is able to detect, decode, and separate all these information sources. Understanding this ability and emulating it on a machine has been an important but elusive scientific and engineering goal for a long time. This talk critically surveys the situation in the speech recognition field. It puts automatic recognition of speech in perspective with other acoustic signal detection and classification tasks, reviews some historical, contemporary, and evolving techniques for machine recognition of speech, critically compares competing techniques, and gives some examples of applications in speech, speaker, and language recognition and identification. The talk is intended for an audience interested but not directly involved in the processing of speech.

Automatic Classification of Acoustic Emission Patterns

Abstract: The problem of automatic classification of acoustic emission signals using techniques derived from pattern recognition is addressed in this paper. The data were taken from laboratory experimental work on a box girder of a bridge in which the acoustic emission (AE) generation mechanism and location were monitored. Two statistical methods and a neural network procedure have been used to classify the data into groups representing different AE generation mechanisms. The classifiers are constructed using the traditional AE features – four parameters from each burst. Principal component analysis is used to reduce the dimension of the AE data feature vectors to two dimensions, resulting in simple visualisations of the data.

Multi-wavelet detection and denoising of low-frequency chirp signals using adaptive wavelet methods

The detection and classification of underwater acoustic signals embedded in noise is a fundamental problem of interest to the signal processing community. The use of wavelet transforms is a recent development in digital signal processing which has been applied in many different areas. A particular type of wavelet is the chirplet, which includes frequency variation as well as time shift and scaling. Both linear and polynomial chirp signals are present in underwater acoustic signals generated by such sources as biologics, ships, and submarines. Distinguishing the features of these chirps relative to other ambient noise shows promise as an initial step in classification of underwater acoustic signals. Removal of unwanted broadband signal components via wavelet methods has been shown to outperform other noise removal processes such as low-pass and high-pass filtering and Weiner filtering. Examples of low-frequency simulated chirp signals with additive noise have been generated. A multi-wavelet packet method for detection and denoising low-frequency signals containing multiple chirps embedded in noise using a specific wavelet designed for polynomial chirp signals is shown. [Research supported in part by ONR.]

An efficient fuzzy based technique for signal classification

Fuzzy systems are currently finding practical applications, ranging from soft regulatory control in consumer products to accurate modelling of non-linear systems. This paper presents the design of a classification system for vehicle acoustic signal classification. Traffic management and information systems rely on a suite of sensors for estimating traffic parameters. Currently inductive loop detectors and video-based systems are often used to count and detect vehicles. Loop detectors are expensive to maintain and video-based systems are sensitive to environmental conditions and do not perform well in vehicle classification. Vehicle classification is important in the computation of the percentages of vehicle classes that use streets and motorways. The use of an automated system can lead to adequate road surface maintenance with obvious results in cost and quality. However the sound of a working vehicle could provide an important clue to the vehicle type. A novel approach, based on adaptive fuzzy logic systems, has been discussed in this paper. Its performance is evaluated through a simulation study, using metered data collected from a roadside microphone-array sensor at the Valle d'Aosta highway in north-western Italy. The results indicate that the fuzzy classifier based on the proposed defuzzification method, namely area of balance (AOB), provide more accurate classifications compared to other classifiers.

Wavelet detection and denoising of low-frequency chirp signals

The detection and classification of underwater acoustic signals embedded in noise is a fundamental problem of interest to the Navy. The use of wavelet transforms is a recent development in digital signal processing that has been applied in many different areas. A particular type of wavelet is the chirplet, which includes frequency variation as well as time shift and scaling. The analysis of low-frequency signals containing multiple chirps using wavelet and chirplet techniques is demonstrated. Examples of low-frequency synthetic chirp signals are generated. Denoising and feature extraction of these signals using various wavelets with wavelet packet techniques are described.

A Wavelet Packet Algorithm for Classification and Detectionof Moving Vehicles

In this paper we propose a robust algorithm that solves two related problems: 1) Classification of acoustic signals emitted by different moving vehicles. The recorded signals have to be assigned to pre-existing categories independently from the recording surrounding conditions. 2) Detection of the presence of a vehicle in a certain class via analysis of its acoustic signature against the existing database of recorded and processed acoustic signals. To achieve this detection with practically no false alarms we construct the acoustic signature of a certain vehicle using the distribution of the energies among blocks which consist of wavelet packet coefficients. We allow no false alarms in the detection even under severe conditions; for example when the acoustic recording of target object is a superposition of the acoustics emitted from other vehicles that belong to other classes. The proposed algorithm is robust even under severe noise and a range of rough surrounding conditions. This technology, which has many algorithmic variations, can be used to solve a wide range of classification and detection problems which are based on acoustic processing which are not related to vehicles. These have numerous applications.

Automatic classification of wideband acoustic signals

Until now, very few contributions were published in the field of wideband acoustic signal recognition, especially for handling impulsive noise signals such as glass breaking, detonations, or door slams, as encountered in security applications, where the signals are highly nonstationary and composed of higher frequency components. This paper shows how the audio alarm recognition problem can efficiently be tackled using either pattern recognition methods relying on Bayes classifiers, or on artificial neural networks (ANN). After extraction of filterbank coefficients in the acoustic analysis module, typical feature vectors are achieved from the concatenation of k consecutive signal frames, in order to exploit dynamic temporal information. The redundancy induced by this dynamic modeling requires a reduction of the feature space dimension, performed with a conventional principal component analysis. The performance of both systems is evaluated experimentally (7000 tests) for the classification of three types of noise events (glass breaking, door slams, and stationary noises). The score of correct classification reaches 99% for the statistical approach, and 98% for the ANN-based system. It is further noticed that errors do not occur on the same signal files for both methods. Thus robustness enhancement can be reasonably expected, when using hybrid methods or fusion strategies.

Maximum likelihood approach to image texture and acoustic signal classification

The authors describe a method of classifying natural textures based on the maximum likelihood parameter estimation technique. The novelty of the technique lies in the use of textural features that are derived from the subbands of a wavelet transformed image via the co-occurrence matrices. A maximum likelihood classifier is designed using a set of training texture samples. Ten different Brodotz textures have been classified using this procedure with an average classification accuracy of 99.7%. The main emphasis is to apply this technique to the classification of underwater acoustic signals. A time–frequency plot is obtained for each segment of the acoustic signal and then converted to an intensity pattern. The textural classification scheme is then applied to the intensity patterns of the acoustic signals. Eight different underwater acoustic signals have been classified by this procedure with an average accuracy of 99.99%.

Classification of mixed acoustic emission signals via neural networks : Yang, J.; Dumont, G.A. 3rd Annual International Conference on Applications of Neural Networks, Orlando, Florida (United States), 21–24 Apr. 1992. pp. 849–854. SPIE (1992) ISBN 0819408743

Classification of mixed acoustic emission signals via neural networks : Yang, J.; Dumont, G.A. 3rd Annual International Conference on Applications of Neural Networks, Orlando, Florida (United States), 21–24 Apr. 1992. pp. 849–854. SPIE (1992) ISBN 0819408743

A rules-based approach to classification of chemical acoustic emission signals

Abstract Soulsbury, K.A., Wade, A.P., and Sibbald, D.B., 1992. A rules-based approach to classification of chemical acoustic emission signals. Chemometrics and Intelligent Laboratory Systems, 15: 87–105 Chemical acoustic emission studies pose unique data analysis problems, since a single experiment can result in capture of several thousand signals, each of at least 1 Kb in length, at rates of more than 10/s. In studies where waveform analysis is considered valuable, automated methods for data reduction and information extraction are therefore essential. This paper presents a rules-based approach to the automated classification of chemical acoustic emission signals. Signal classification rules are generated by analysis of data from preliminary experiments which form a training set. Rule clauses indicate the expected values for the descriptive statistical factors (descriptors) which best characterize signals as belonging to recognized classes (e.g. gas evolution, crystallization, background noise). The rules may then be applied to classify signals in further data sets. Modification of the rules to include only those descriptors which best differentiate the classes of interest provides optimal performance. This approach has several advantages: since only the signal class, time of acquisition (and possibly descriptor values) need to be stored, long-term data storage is greatly decreased, whilst retaining analytical information; automated rejection of signals due to background noise results in sensitivity improvements, since lower trigger levels may be used; furthermore, the improved selectivity provides greater confidence that signals captured can be attributed to processes of interest. Thus, the approach opens the door to intelligent, real-time data reduction, more accurate measurements of emission rates, and automatic identification of dominant physicochemical emission mechanisms. Once trained, the automated signal classifier performed as well as or better than human pattern recognition, and required only a fraction of the time. To illustrate the utility of the approach, acoustic emission waveform characterization rules were obtained for and applied to three chemical systems. For pyrolysis of poly(vinyl chloride), noise signals were effectively discriminated from signals from the chemical process. Rules for copper(II) sulfate recrystallization resulted in improved certainty in determining the time of the onset and the rate of crystal fracture. In hydration of silica gel, rules automatically classified signals as being from fracture or gas evolution.

Artificial neural networks for feature extraction from acoustic emission signals : Colloquium on Measurements, Modelling and Imaging for Non-Destructive Testing, London (United Kingdom), 27 Mar. 1991. pp. 6/1–6/6. IEE (1991) Digest No. 1991/054

The authors discuss the use of artificial neural network (ANN) techniques to carry out the task of feature classification of acoustic emission signals. A series of experiments aimed at determining the quality of feature extraction and the resulting ANN classifier performance have been performed. The performance assessment was based on the time required to train the network given the same training and testing input vectors set.

A methodology for detection and classification of some underwater acoustic signals using time-frequency analysis techniques

Signal detection techniques based on time-frequency signal analysis with the Wigner-Ville distribution (WVD) and the cross Wigner-Ville distribution (XWVD) are presented. These techniques are shown to provide high-resolution signal characterization in a time-frequency space, and good noise rejection performance. This type of detection is applied to the signaturing, detection, and classification of specific machine sounds: the individual cylinder firings of a marine engine. For this task, a four-step procedure has been devised. The autocorrelation function (ACF) is first used to ascertain the number of engine cylinders and the firing rate of the engine. Further correlation techniques are then used to detect the time at which individual cylinder firing events occur. WVD- and XWVD-based analyses follow to procedure high-resolution time-frequency signatures. Finally, two-dimensional correlations are used for the classification of the individual cylinders. The proposed methodology is tested on real data. XWVD-based detection is also applied to detection of a transient with unknown waveshape (using real data). >

Interpretation of crack propagation events on the basis of acoustic emission signals and other non-destructive test results concerning cracks in a RPV nozzle under cyclic thermal shock loading

Cyclic thermal shock tests have been performed on a RPV nozzle (HDR) with cracks in order to apply and to assess technics available for the description of real cracks (non-destructive testing, fracture mechanics calculation). In this paper the application of dye penetrant testing, potential drop testing and acoustic emission measurements on the nozzle is considered. Some experimental results are described in detail; assessments have been possible on the basis of destructive testing (fractography). The determination of crack parameters for correlating these NDT results with fracture mechanics data are discussed such as crack depth, a, cracklength, crack area and crack growth. Crack growth is considered over a period of cycles, a( N), per cycle, d a/d N, and during a cycle, d a/d t. A selection of numerical NDT findings is presented and compared with the results of fracture mechanics calculations. The investigation leads to the following conclusions: 1. 1. The crack dimensions found by NDT are not quite conservative (fractography). 2. 2. As the performed fracture mechanics calculations overestimated the crack growth advanced NDT data like d a/d N indicate starting points for improvements of the calculations (influence of mean stress). 3. 3. Although acoustic emission measurements give no quantitative fracture mechanics parameters qualitative correlations yield decisive understanding of dominating crack growth mechanisms (corrosion). 4. 4. Reversely different comparisons between acoustic emission and fracture mechanics results seem to confirm the classification of acoustic emission signals (crack surface friction, crack growth).

Game theoretical signal classification: Application to imperfect or non-cooperative learning

This paper studies in theory, and gives a solution to, the following concerns: (1) obtain alternative classification decisions, ranked by some decreasing order of class membership probabilities; (2) imperfect teacher at the learning stage, or effect of labelling errors due to unsupervised learning; (3) non-cooperative teacher manipulating the a-priori class probabilities; (4) unknown a-priori class probabilities. This is carried out by considering a game between the recognition system and the teacher, in a game theoretical framework. Both players will ultimately select “mixed strategies” which are probability distributions over the set of N pattern classes. Within the context of signal classification, these N classes are N alternate signal classes e.g., sources, modelled by autoregressive processes. This approach, and the concerns (1)–(4) are especially relevant for the performance enhancement of a number of acoustic signal classification systems (e.g., seismic exploration, intrusion detection, sonar).

Discrimination of natural vocalizations by budgerigars (Melopsittacus undulatus) using operant techniques

Budgerigars (Melopsittacus undulatus) were trained by operant techniques on a GO‐NOGO discrimination task requiring the classification of complex species specific acoustic signals. Budgerigars began training on one pair of contact calls. After reaching a criterion of 85% correct, a third stimulus was added to the classification task. This procedure was reported each time budgerigars reached criterion on the current set of calls. By 70 sessions (days) budgerigars demonstrated the capacity for correctly identifying/classifying ten contact calls. Responses to probe stimuli indicate that the budgerigars were relying on the features of each call, not a rule or rules, in making a correct classification. Canaries (Serinus canarius) trained on the identical task failed to reach criterion on a single pair of budgerigar contact calls. These data suggest species differences in the perception and learning of complex vocalizations. [Work supported by NINCDS and NICHHD.]

On the classification of underwater acoustic signals. I. An environmentally adaptive approach

Environmental conditions and geometries have a pronounced effect upon signals received as scattered underwater acoustic waveforms. Information which may be processed to yield accurate assessments of the current environmental condition (and geometry) is important to the effective analysis and classification of these received signals. However, such environmental information is often ignored, or at best introduced indirectly, in degradation, through the usual formulation of composite hypothesis tests. Here, a methodology is developed which permits one to introduce estimates of the current environmental state (i.e., certain environments and object properties) in order to adapt the classification algorithm to the situation. The implications of such an environmentally adaptive approach are discussed, and general procedures for estimation, feature extraction, and feature selection are then presented. The pertinent features of analytic statistical–physical models of the medium provide useful guides, not only for the type of experiments needed to define these environmental states, but also for models, for deriving ’’artificial’’ data, or waveforms, by which the adaptive portions of the classification process can be studied and implemented.

On the classification of underwater acoustic signals. II. Experimental applications involving fish

In a previous paper (Part I) a general approach to adaptive classification has been presented. That approach is applied here (Part II) to classification of waveforms obtained in a number of illustrative experiments. Experiments at 116 kHz (in the ARL Test Tank Facility) have provided backscattered waveforms (echoes) from known types of scatterers, randomly located and with random acoustic cross sections. Scatterers of primary interest here are a variety of small fish. Signals which form competing categories are scattering from seaweed (aquarium grass), surface reverberation, and artifacts (glass bottles). The data obtained are representative of three classification problems that have some practical significance in underwater acoustics. These three problems require distinguishing between the received signals which result from scattering by: (1) fish versus air–water surface, seaweed, and/or artifacts; (2) fish and air–water surface versus air–water surface only; and (3) ’’low density’’ versus ’’medium density’’ versus ’’high density’’ of fish. In each case the waveforms are analyzed and features appropriate to the classification task are extracted from a portion of the data. These features are then used to design a classification algorithm for the environmentally adaptive approach and for its conventional counterpart. These algorithms are then used on the remaining data to make classifications. The superiority and usefulness of the proposed environmentally adaptive approach is demonstrated here by the noticeable improvement in quantitative classification which results from the exploitation of environmental state data.

On the classification of underwater acoustic signals. I. An environmentally adaptive approach

Environmental conditions and geometries have a pronounced effect upon the signals received as scattered underwater acoustic waveforms. The use of these environmental factors is important. However, these factors are often ignored or at best are introduced indirectly in the formulation of composite hypothesis testing when analyzing the received signals and classifying causes of the scattering. Here we develop a methodology which permits one to introduce estimates of the current environmental state (i.e., certain environments and geometries) in order to adapt the classification algorithm to the situation. The pertinent features of statistical-physical models of the medium (as proposed by Middleton, for example) are useful guides to the experiments needed for the definition of these environmental states. The implications of an environmentally adaptive approach are discussed, and procedures for estimation, feature extraction, and feature selection are presented. [Work supported by the Naval Sea Systems Command.]