Results 1  10
of
10
Optimizing TimeFrequency Kernels for Classification
, 2001
"... In many pattern recognition applications, features are traditionally extracted from standard timefrequency representations (TFRs). This assumes that the implicit smoothing of, say, a spectrogram is appropriate for the classification task. Making such assumptions may degrade classification performa ..."
Abstract

Cited by 12 (1 self)
 Add to MetaCart
In many pattern recognition applications, features are traditionally extracted from standard timefrequency representations (TFRs). This assumes that the implicit smoothing of, say, a spectrogram is appropriate for the classification task. Making such assumptions may degrade classification performance. In general, any timefrequency classification technique that uses a singular quadratic TFR (e.g., the spectrogram) as a source of features will never surpass the performance of the same technique using a regular quadratic TFR (e.g., Rihaczek or WignerVille). Any TFR that is not regular is said to be singular. Use of a singular quadratic TFR implicitly discards information without explicitly determining if it is germane to the classification task. We propose smoothing regular quadratic TFRs to retain only that information that is essential for classification. We call the resulting quadratic TFRs classdependent TFRs. This approach makes no a priori assumptions about the amount and type of timefrequency smoothing required for classification. The performance of our approach is demonstrated on simulated and real data. The simulated study indicates that the performance can approach the Bayes optimal classifier. The realworld pilot studies involved helicopter fault diagnosis and radar transmitter identification.
Improved Optimization of TimeFrequency Based Signal Classifiers
, 2001
"... TimeFrequency Pepresentations (TFPs) are efficient tools for nonstationary signal classification. However, the choice of the TFP and of the distance measure employed is critical when no prior information other than a learning set of limited size is available. In this letter, we propose to jointly o ..."
Abstract

Cited by 8 (4 self)
 Add to MetaCart
TimeFrequency Pepresentations (TFPs) are efficient tools for nonstationary signal classification. However, the choice of the TFP and of the distance measure employed is critical when no prior information other than a learning set of limited size is available. In this letter, we propose to jointly optimize the TFP and distance mea sure by minimizing the (estimated) probability of classifi cation error. The resulting optimized classification method is applied to multicomponent chirp signals and real speech records (speaker recognition). Extensive simulations show the substantial improvement of classification performance obtained with our optimization method.
Optimal selection of timefrequency representations for signal classification: a kerneltarget alignment approach
 in Proc. IEEE International Conference on Acoustics, Speech and Signal Processing
, 2006
"... In this paper, we propose a method for selecting timefrequency distributions appropriate for given learning tasks. It is based on a criterion that has recently emerged from the machine learning literature: the kerneltarget alignment. This criterion makes possible to find the optimal representation ..."
Abstract

Cited by 4 (4 self)
 Add to MetaCart
(Show Context)
In this paper, we propose a method for selecting timefrequency distributions appropriate for given learning tasks. It is based on a criterion that has recently emerged from the machine learning literature: the kerneltarget alignment. This criterion makes possible to find the optimal representation for a given classification problem without designing the classifier itself. Some possible applications of our framework are discussed. The first one provides a computationally attractive way of adjusting the free parameters of a distribution to improve classification performance. The second one is related to the selection, from a set of candidates, of the distribution that best facilitates a classification task. The last one addresses the problem of optimally combining several distributions.
Using Optimized TimeFrequency Representations for Acoustic Quality Control of Motors
"... In this article a problem of industrial quality control is investigated. Electric motors have to be tested for proper working by analyzing their emitted noise signal in regular running. The noise signal being (approximately) periodic, it can be seen as a sequence of single signal segments p i (t); ..."
Abstract

Cited by 1 (1 self)
 Add to MetaCart
In this article a problem of industrial quality control is investigated. Electric motors have to be tested for proper working by analyzing their emitted noise signal in regular running. The noise signal being (approximately) periodic, it can be seen as a sequence of single signal segments p i (t); i = 1; : : : ; N ; t = 1; : : : ; L i with approximately equal length L i . These segments are represented in the timefrequency space by their timefrequency representations (TFR) C i (t; ; \Phi). For each of the two considered fault classes "beating" and "gear noise" a fault parameter is defined based on the sequence of periods. This parameter implements a coarse knowledge of the corresponding physical fault. Since the representation of the signal segments depends explicitely on the kernel function \Phi of the TFR, the ability of the fault parameter to discriminate between good and defective motors can be optimized by varying \Phi. The resulting optimum kernel functions enables a reliable c...
Classification of induction machine faults by optimal timefrequency representations
 IEEE Trans. Ind. Electron. 2008
"... Abstract—This paper presents a new diagnosis method of induction motor faults based on time–frequency classification of the current waveforms. This method is based on a representation space, a selection criterion, and a decision criterion. In order to define the representation space, an optimized ti ..."
Abstract

Cited by 1 (0 self)
 Add to MetaCart
Abstract—This paper presents a new diagnosis method of induction motor faults based on time–frequency classification of the current waveforms. This method is based on a representation space, a selection criterion, and a decision criterion. In order to define the representation space, an optimized time–frequency representation (TFR) is designed from the time–frequency ambiguity plane. The selection criterion is based on Fisher’s discriminant ratio, which allows one to maximize the separability between classes representing different faults. A distinct TFR is designed for each class. The following two classifiers were used for decision criteria: the Mahalanobis distance and the hidden Markov model. The flexibility of this method allows an accurate classification independent from the level of load. This method is validated on a 5.5kW induction motor test bench. Index Terms—Diagnosis, hidden Markov model (HMM), induction motor, time–frequency classification. I.
Nonstationary signal analysis with kernel machines
"... This chapter introduces machine learning for nonstationary signal analysis and classification. It argues that machine learning based on the theory of reproducing kernels can be extended to nonstationary signal analysis and classification. The authors show that some specific reproducing kernels allow ..."
Abstract
 Add to MetaCart
(Show Context)
This chapter introduces machine learning for nonstationary signal analysis and classification. It argues that machine learning based on the theory of reproducing kernels can be extended to nonstationary signal analysis and classification. The authors show that some specific reproducing kernels allow pattern recognition algorithm to operate in the timefrequency domain. Furthermore, the authors study the selection of the reproducing kernel for a nonstationary signal classification problem. For this purpose, the kerneltarget alignment as a selection criterion is investigated, yielding the optimal timefrequency representation for a given classification problem. These links offer new perspectives in the field of nonstationary signal analysis, which can benefit from recent developments of Timefrequency and timescale distributions have become increasingly popular tools for analysis and processing of nonstationary signals. These tools map a onedimensional signal into a twodimensional distribution, a function of both time
CLASSIFICATION IN THE GABOR TIME–FREQUENCY DOMAIN OF NON–STATIONARY SIGNALS EMBEDDED IN HEAVY NOISE WITH UNKNOWN STATISTICAL DISTRIBUTION
"... A new supervised classification algorithm of a heavily distorted pattern (shape) obtained from noisy observations of nonstationary signals is proposed in the paper. Based on the Gabor transform of 1D nonstationary signals, 2D shapes of signals are formulated and the classification formula is deve ..."
Abstract
 Add to MetaCart
(Show Context)
A new supervised classification algorithm of a heavily distorted pattern (shape) obtained from noisy observations of nonstationary signals is proposed in the paper. Based on the Gabor transform of 1D nonstationary signals, 2D shapes of signals are formulated and the classification formula is developed using the pattern matching idea, which is the simplest case of a pattern recognition task. In the pattern matching problem, where a set of known patterns creates predefined classes, classification relies on assigning the examined pattern to one of the classes. Classical formulation of a Bayes decision rule requires aprioriknowledge about statistical features characterising each class, which are rarely known in practice. In the proposed algorithm, the necessity of the statistical approach is avoided, especially since the probability distribution of noise is unknown. In the algorithm, the concept of discriminant functions, represented by Frobenius inner products, is used. The classification rule relies on the choice of the class corresponding to the max discriminant function. Computer simulation results are given to demonstrate the effectiveness of the new classification algorithm. It is shown that the proposed approach is able to correctly classify signals which are embedded in noise with a very low SNR ratio. One of the goals here is to develop a pattern recognition algorithm as the best possible way to automatically make decisions. All simulations have been performed in Matlab. The proposed algorithm can be applied to nonstationary frequency modulated signal classification and nonstationary signal recognition.
Classification of Time Series With Optimized TimeFrequency Representations
"... this article we consider the problem of classification of finite time series of length L: f(t); t = 1; : : : ; L, for the case of two classes. The usual way of dealing with this problem is the following: First a number of features ..."
Abstract
 Add to MetaCart
this article we consider the problem of classification of finite time series of length L: f(t); t = 1; : : : ; L, for the case of two classes. The usual way of dealing with this problem is the following: First a number of features
Online diagnosis of induction motor faults
, 2007
"... Abstract: A new method of automatic diagnosis of induction motor faults based on the timefrequency ambiguity plane analysis of the current waveforms. This method is composed of two sequential processes: a feature extraction and a classification. In the process features extraction, the timefrequenc ..."
Abstract
 Add to MetaCart
(Show Context)
Abstract: A new method of automatic diagnosis of induction motor faults based on the timefrequency ambiguity plane analysis of the current waveforms. This method is composed of two sequential processes: a feature extraction and a classification. In the process features extraction, the timefrequency representation (TFR) have been designed for maximizing the separability between classes representing the different faults; bearing fault, stator fault and rotor fault. The classification of a new signal is based on the Mahalanobis distance. Key words: automatic diagnosis, timefrequency representation, bearing fault, stator fault, rotor fault