Wyniki wyszukiwania - Biblioteka Nauki

1

Statistical testing of segment homogeneity in classification of piecewise-regular objects

100%

Savchenko A. , Belova N.

|

tom 25

|

nr 4

915-925

EN

The paper is focused on the problem of multi-class classification of composite (piecewise-regular) objects (e.g., speech signals, complex images, etc.). We propose a mathematical model of composite object representation as a sequence of independent segments. Each segment is represented as a random sample of independent identically distributed feature vectors. Based on this model and a statistical approach, we reduce the task to a problem of composite hypothesis testing of segment homogeneity. Several nearest-neighbor criteria are implemented, and for some of them the well-known special cases (e.g., the Kullback-Leibler minimum information discrimination principle, the probabilistic neural network) are highlighted. It is experimentally shown that the proposed approach improves the accuracy when compared with contemporary classifiers.

2

Stable or not stable? Recognizing surge based on the pressure signal

88%

Garcia D. , Liśkiewicz G.

Transactions of the Institute of Fluid-Flow Machinery

|

2016

|

tom nr 133

55--68

EN

The surge protection may be worth millions of dollars. This is typical price of a centrifugal compressor repair combined with additional cost of nonfunctionality of an industry employing it. This threat is normally secured by application of an antisurge systems. Typically they are activated at predeﬁned working conditions when compressor mass ﬂow rate approaches region aﬀected by the surge. As a result those systems are vastly limiting its operational range usually by a desirable region where compressor attains large pressure ratio. Therefore, a modern antisurge systems are aiming at diminishing this tradeoﬀ by reacting to the real pressure signal gathered at high frequency. This paper presents one of those methods employing singular spectrum analysis. This algorithm has not been widely used for this application, while it was shown herein that it may bring clear distinction between stable and nonstable working condition, even at presurge conditions. Hence in further perspective it may bring anti-surge protection quality, that was not met with another methods.

3

A parallel net of (1-NN, K-NN) classifiers for optical inspection of surface defects in ferrites

75%

Jóźwik A. , Chmielewski L. , Skłodowski M. , Cudny W.

Machine Graphics and Vision

|

1998

|

tom Vol. 7, No. 1/2

99-112

EN

The challenging task of optical inspection of surface defects in ferrite cores has been successfully approached with a set of methods. In this paper the attention is paid to the k Nearest Neighbours classifier developed for the system. A parallel net of two-decision classifiers is presented. The combination of the 1-NN and k-NN rules reduces the training time. A great part of computations is restricted to the class overlap area. The classification quality is significantly improved if a separate feature selection for each of the component classifiers is done. A dramatic improvement of classification speed obtained by reference patterns sets reduction for component classifiers is vital, as in the considered task the classifier is used for recognition of pixels. The proposed modifications of the classifier are of general usefulness for pattern recognition. The presented quality inspection system can be applied to various defect detection tasks.

4

Statistical testing of segment homogeneity in classification of piecewise-regular objects

75%

Savchenko A. V. , Belova N. S.

International Journal of Applied Mathematics and Computer Science

|

2015

|

tom Vol. 25, no. 4

915--925

EN

The paper is focused on the problem of multi-class classification of composite (piecewise-regular) objects (e.g., speech signals, complex images, etc.). We propose a mathematical model of composite object representation as a sequence of independent segments. Each segment is represented as a random sample of independent identically distributed feature vectors. Based on this model and a statistical approach, we reduce the task to a problem of composite hypothesis testing of segment homogeneity. Several nearest-neighbor criteria are implemented, and for some of them the well-known special cases (e.g., the Kullback–Leibler minimum information discrimination principle, the probabilistic neural network) are highlighted. It is experimentally shown that the proposed approach improves the accuracy when compared with contemporary classifiers.

5

Ventilatory response to hypoxia in experimental pathology of the diaphragm

63%

Sokolowska B. , Jozwik A.

|

nr 3

121-126

EN

In this study, we examined the usefulness of arterial blood gas variables, as changed by the hypoxic stimulus, in discerning various experimentally-induced conditions of diaphragm weakness in anesthetized cats. We defined three experimental situations (models): (i) intact muscle, statistical Class I, (ii) four degrees of muscle dysfunction (after sequential diaphragm denervation), Classes II-V, and (iii) entirely paralyzed muscle, Class VI. Responses to a hypoxic stimulus in the above-mentioned conditions were evaluated by using the methods of the pattern recognition theory. We found that before the hypoxic stimulus, with partial but of different severity denervation of the diaphragm, the k-nearest neighbor classifier (k-NN) assigned 100% of the classified cases to Class II (one phrenic nerve rootlet cut). In contrast, during hypoxia only 67% of cases were assigned to Class II, the remaining being spread throughout other classes of muscle weakness. When one limits the procedure to the extreme classes: Class I (intact diaphragm) and Class VI (totally denervated diaphragm), the k-NN picks out 33% and 50% cases of bilateral diaphragm paralysis before and during hypoxia, respectively. We conclude that any remaining innervations of the diaphragm ensure the functionally optimal level of lung ventilation that may waver when hypoxia develops.

6

Dynamical system analysis of unstable flow phenomena in centrifugal blower

63%

Garcia D. , Stickland M. , Liśkiewicz G.

Zeszyty Naukowe. Cieplne Maszyny Przepływowe - Turbomachinery / Politechnika Łódzka

|

2014

|

tom nr 145

43--46

EN

Methods of dynamical system analysis were employed to analyze unsteady phenomena in the centrifugal blower. Pressure signals gathered at different control points were decomposed into their Principal Components (PCs) by means of Singular Spectrum Analysis (SSA). Certain number of PCs was considered in the analysis based on their statistical correlation. Projection of the original signal onto its PCs allowed to draw the phase trajectory that clearly separated non-stable blower working conditions from its regular operation.