Showing papers on "Mahalanobis distance published in 1989"

A practical method for outlier detection in autoregressive time series modelling

Abstract: A practical method is developed for outlier detection in autoregressive modelling. It has the interpretation of a Mahalanobis distance function and requires minimal additional computation once a model is fitted. It can be of use to detect both innovation outliers and additive outliers. Both simulated data and real data re used for illustration, including one data set from water resources.

A program for calculating Mahalanobis distances using principal component analysis

Abstract: A technique for pattern recognition analysis of near-infrared reflectance spectra is described. Classification of samples is achieved by using the Mahalanobis distances of spectra in a principal component subspace. Probability levels for class membership are determined from the Chi-squared distribution.

On the inadmssibelity of ridge estimator in a linear model

Abstract: The necessary and sufficient conditions for the inadmissibility of the ridge regression is discussed under two different criteria, namely, average loss and Pitman nearness. Although the two criteria are very different, same conclusions are obtained. The loss functions considered in this article are th likelihood loss function and the Mahalanobis loss function. The two loss functions are motivated from the point of view of classification of two normal populations. Under the Mahalanobis loss it is demonstrated that the ridge regression is always inadmissible as long as the errors are assumed to be symmetrically distributed about the origin.

The studentized location linear discriminant function

Abstract: The location linear discriminant function is used in a two-population classification problem when the available data are generated from both binary and continuous random variables. Asymptotic distribution of the studentized location linear discriminant function is derived directly without the inversion of the corresponding characteristic function. The resulting plug-in estimate of the overall error of misclassification consists of the estimate based on the limiting distribution of the discriminant plus a correction term up to the second order. By comparison, our estimate avoids exact knowledge of the Mahalanobis distances which is necessary when the expansions of Vlachonikolis (1985) are used in the case of an arbitrary cut-off point. An example is re-examined and analysed in the present context.

Artifact detection in sleep EEG recording

Abstract: An investigation was conducted with the aim of achieving a completely automatic detection of artifacts resulting from muscular activity and ocular movements superimposed on an electroencephalogram (EEG) signal. Muscular artifacts were detected by a pattern-recognition approach; different features were tested to achieve reliable and real-time results. Pattern classes were constructed based on selected features and using the K-means clustering algorithm. Fisher's linear discriminant, Mahalanobis distance, and the Q-NN rule were the methods tested within the task of supervised classification, using the pattern classes found in the clustering step. Ocular artifacts were detected by an algorithm which compares simultaneously recorded EOG (electrooculogram) and EEG signals. Whenever a significant correlation between them is found, an ocular artifact is detected. If one or both of those artifact types are detected, the actual segment is replaced by an EEG simulation using an AR (autoregressive) modeled system driven by Gaussian white noise at its input. >

Amendments and Corrections: The Mahalanobis Distance and Elliptic Distributions

Abstract: On page 518 the expressions for u2 and v2 should be U2= S2 _ (n3/N2), v2=t72-(m3/2N2). In view of the complexity of the formula (3d 1), it is worth noting that it gives the same numerical results, in a number of examples, as the subsequent formula of Mielke & Berry (1985). It also agrees numerically in these examples with the exact third central moment calculated using the enumeration algorithm of Saunders referred to in our paper.

On a Classification Problem: Ranking and Selection Approach

Abstract: : This paper deals with a classification problem based on ranking and selection approach. We assume that the populations follow multivariate normal distribution. The corresponding selection problem is to choose the population with the smallest Mahalanobis distance. The subset selection approach is considered throughout this paper. Sometimes the indifference zone approach is also proposed. It should be pointed out that, for the subset selection approach, we need not assume that the individual to be classified belongs to one of the several given categories. Keywords: Classification rules; Multivariate normal populations; Mahalanobis distance; Ranking and selection; Probability of correct classification.

An investigation of speaker verification accuracy using fundamental frequency and duration as distinguishing features

Abstract: A method of automatically verifying a person's identity based on measurements of features taken from a sample of speech is presented. The design goal was to develop a simple, efficiently implemented system, compared with existing systems, while retaining as much accuracy as possible. Parameters extracted from the utterance are fundamental frequency of vocal cord vibration and duration of voicing. Mahalanobis and Euclidean distances are used to compare the sets of features. An accuracy of 100% was obtained for one individual with the Mahalanobis distance measure; several other individuals had performance accuracies over 90%. On the whole, accuracy rates achieved in the simulation were about 80%. >

Signed euclidean distance transform applied to shape analysis

Abstract: The signed Euclidean distance transform is a modified version of Danielsson’s Euclidean distance transform [1]. This distance transform produces a distance map in which each pixel is a vector of two integer components. If a distance map is created inside the objects, the two integer values of a pixel in the distance map represent the displacements of the pixel from the nearest background point in x and y directions, respectively. The unique feature of this distance transform that a vector in the distance map is always pointing to the nearest background point is exploited in several applications, such as the detection of dominant points in digital curves, curve smoothing, computing Dirichlet tessellations and finding convex hulls.

Use of robust estimators in parametric classifiers

Abstract: The authors examine several robust estimators of the mean and covariance matrix and their effect on the probability of error in classification. A simple and intuitive iterative approach which weighs each data point as a function of its distance (Mahalanobis distance) is given in the proposed algorithm, which showed good convergence behavior in simulation with artificial data. Some comments on alpha -ranked ( alpha -trimmed) estimators are presented. >