Wyniki wyszukiwania - BazTech

1

Spectral entropy and deep convolutional neural network for ECG beat classification

Asgharzadeh-Bonab Akbar, Amirani Mehdi Chehel, Mehri Alaeddin

Biocybernetics and Biomedical Engineering

|

2020

|

Vol. 40, no. 2

691--700

EN

Sudden cardiac death is the result of abnormal heart conditions. Therefore, early detection of such abnormal conditions is vital to identify heart problems. Hence, in this paper, we aim to present a new computer-aided diagnosis (CAD) method based on time-frequency analysis of electrocardiogram (ECG) signals and deep neural networks for arrhythmia detection. Time-frequency transforms have the capability of providing spectral information at different times, which is very useful for analyzing non-stationary signals. On the other side, entropy is an attractive measurement from ECG signals which can distinguish different types of them. In this paper, time-frequency spectral entropy is proposed to extract the efficient features from ECG signals. All computed entropies cannot provide separability among different classes, two-directional two-dimensional principal component analysis (2D2PCA) can be used to reduce the dimension of the extracted features. Finally, the convolutional neural network (CNN) classifies the time-frequency features to diagnose the ECG beat signals and detect arrhythmias. The results show that the spectral entropy can provide good separation between different among ECG beats and the proposed method outperforms the recently introduced method for analyzing ECG signals.

2

A fast neural network learning algorithm with approximate singular value decomposition

Jankowski Norbert, Linowiecki Rafał

International Journal of Applied Mathematics and Computer Science

|

2019

|

Vol. 29, no. 3

581--594

EN

The learning of neural networks is becoming more and more important. Researchers have constructed dozens of learning algorithms, but it is still necessary to develop faster, more flexible, or more accurate learning algorithms. With fast learning we can examine more learning scenarios for a given problem, especially in the case of meta-learning. In this article we focus on the construction of a much faster learning algorithm and its modifications, especially for nonlinear versions of neural networks. The main idea of this algorithm lies in the usage of fast approximation of the Moore–Penrose pseudo-inverse matrix. The complexity of the original singular value decomposition algorithm is O(mn2). We consider algorithms with a complexity of O(mnl), where l < n and l is often significantly smaller than n. Such learning algorithms can be applied to the learning of radial basis function networks, extreme learning machines or deep ELMs, principal component analysis or even missing data imputation.

3

Detection of type-2 diabetes using characteristics of toe photoplethysmogram by applying support vector machine

Nirala Neelamshobha, Periyasamy R., Singh Bikesh Kumar, Kumar Awanish

Biocybernetics and Biomedical Engineering

|

2019

|

Vol. 39, no. 1

38--51

EN

Diabetes mellitus (DM) is one of the most widespread and rapidly growing diseases. With its advancement, DM-related complications are also increasing. We used characteristic features of toe photoplethysmogram for the detection of type-2 DM using support vector machine (SVM). We collected toe PPG signal, from 58 healthy and 83 type-2 DM subjects. From each PPG signal 37 different features were extracted for further classification. To improve the performance of SVM and reduce the noisy data we employed hybrid feature selection technique that reduces the feature set of 37 to 10 on the basis of majority voting. Using 10 selected features set, we gained an accuracy of 97.87%, sensitivity of 98.78% and specificity of 96.61%. Further for the validation of our method we need to do random population test, so that it can be used as a non-invasive screening tool. Photoplethysmogram is an economic, technically easy and completely non-invasive method for both physician and subject. With the high accuracy that we obtained, we hope that our work will help the clinician in screening of diabetes and adopting suitable treatment plan for preventing end organ damage.

4

An unsupervised approach to leak detection and location in water distribution networks

Quiñones-Grueiro M., Verde C., Prieto-Moreno A., Llanes-Santiago O.

International Journal of Applied Mathematics and Computer Science

|

2018

|

Vol. 28, no. 2

283--295

EN

The water loss detection and location problem has received great attention in recent years. In particular, data-driven methods have shown very promising results mainly because they can deal with uncertain data and the variability of models better than model-based methods. The main contribution of this work is an unsupervised approach to leak detection and location in water distribution networks. This approach is based on a zone division of the network, and it only requires data from a normal operation scenario of the pipe network. The proposition combines a periodic transformation and a data vector extension together with principal component analysis of leak detection. A reconstruction-based contribution index is used for determining the leak zone location. The Hanoi distribution network is employed as the case study for illustrating the feasibility of the proposal. Single leaks are emulated with varying outflow magnitudes at all nodes that represent less than 2.5% of the total demand of the network and between 3% and 25% of the node’s demand. All leaks can be detected within the time interval of a day, and the average classification rate obtained is 85.28% by using only data from three pressure sensors.

5

Klasyczna i neuronowa analiza głównych składowych na przykładzie zadania kompresji obrazu

Bartecki K.

Pomiary Automatyka Kontrola

|

2013

|

R. 59, nr 1

34--37

PL

W artykule omówiono zastosowanie analizy składników głównych (PCA) w zadaniu kompresji stratnej sygnału na przykładzie kompresji obrazu. Zadanie zrealizowano z wykorzystaniem klasycznej metody PCA oraz dwóch rodzajów sieci neuronowych: jednokierunkowej, dwuwarstwowej sieci z uczeniem nadzorowanym i jednowarstwowej sieci z uczeniem nienadzorowanym. W każdym z przypadków przeanalizowano wpływ struktury modelu PCA na wartości współczynnika kompresji oraz średniokwadratowego błędu kompresji.

EN

In the paper, lossy data compression techniques based on the principal component analysis (PCA) are considered on the example of image compression. The presented task is performed using the classical PCA method based on the eigen-decomposition of the image covari-ance matrix as well as two different kinds of artificial neural networks. The first neural structure used is a two-layer feed-forward network with supervised learning shown in Fig.1, while the second one is a single-layered network with unsupervised Hebbian learning. In each case considered, the effect of the PCA model structure on the data compression ratio and the mean square reconstruction error is analysed. The compression results for a Hebbian neural network with K=4 PCA units are presented in Figs. 2, 3 and 4. They show that only 4 eigenvectors are able to capture the main features of the processed image, giving as a result high value of the data compression ratio. However, the reconstructed image quality is not sufficient from a practical point of view. Therefore, selection of the appropriate value for K should take into account the tradeoff between a sufficiently high value for the compression ratio and a reasonably low value for the image reconstruction error. The summary results for both classical and neural PCA compression approaches obtained for different number of eigenvectors (neurons) are compared in Fig. 5. The author concludes that a positive aspect of using neural networks as a tool for extracting principal components from the image data is that they do not require calculating the correlation matrix explicitly, as in the case of the classical PCA-based approach.

6

Feature selection of protein structural classification using SVM classifier

Krajewski Z., Tkacz E.

Biocybernetics and Biomedical Engineering

|

2013

|

Vol. 33, no. 1

47--61

EN

Recursive feature elimination method (RFE), cross validation coefficient (CV) and accuracy of classification of test data are applied as a criterion of feature selection in order to find relevant features and to analyze their influence on classifier accuracy. Feature selection method was compared to principal component analysis (PCA) to understand the effectiveness of feature reduction. Support vector machine classifier with radial basis function (RBF) kernel is applied to find the best set of features using grid model selection and to select and assess relevant features. The best selected feature set is then analyzed and interpreted as the source of knowledge about the protein structure and biochemical properties of amino acids included in the protein domain sequence.

7

PCA transformation and Support Vector Machine for recognition of the noisy images

Osowski S., Sikorska-Łukasiewicz K.

Przegląd Elektrotechniczny

|

2012

|

R. 88, nr 3a

4-6

EN

The paper presents the application of principal component analysis (PCA) and Support Vector Machine (SVM) for recognition of face images. PCA is a well known method of optimal reduction of the dimensionality of the vectors, while preserving the most important part of the original information. It maps the N-dimensional original vector x into K-dimensional output vector y, where K

PL

Praca przedstawia zastosowanie transformacji PCA oraz sieci neuronowej SVM do rozpoznawania obrazów twarzy, w szczególności zaszumionych. PCA jest transformacją liniową umożliwiającą optymalną redukcję wymiaru wektora x przy zachowaniu najważniejszej porcji oryginalnej informacji zawartej w tym wektorze. Przy wektorach z dużą zawartością szumu odtworzona informacja jest w dużej mierze odszumiona (szum odpowiada najmniej istotnej części informacji, która podlega obcięciu). Metoda ta dobrze nadaje się do tworzenia cech diagnostycznych, które mogą stanowić sygnały wejściowe dla klasyfikatora SVM dokonującego rozpoznania obrazów. W pracy pokazane zostały wyniki przeprowadzonych eksperymentów rozpoznania twarzy z bazy danych FERET.

8

Modeling and optimization of the feature generator for speaker recognition systems

Majda E., Dobrowolski A. P.

Przegląd Elektrotechniczny

|

2012

|

R. 88, nr 12a

131-136

EN

This paper presents issues related to modeling and optimization of the feature generator for the speaker recognition system (ASR – Automatic Speakers Recognition). The parameterization stage of generating a speech signal (features generation) is fundamental in this type of system because the unique vector of features is crucial in the process of speech recognition. The task is to describe the speech signal using as few descriptors as possible without loss of relevant information for speaker recognition. In addition, the parameterization should be robust to acoustic and technical registration conditions and to the recorded linguistic material. The research presented in this paper focused primarily on the multicriteria optimization of selected parameters of the feature generator based on cepstral analysis that additionally enables the selection of features. Finally, the evaluation of the results was based on the analysis of the main components of a set of descriptors for the samples of voice acquired from 24 speakers.

PL

W referacie przedstawiono zagadnienia związane z modelowaniem i optymalizacją generatora cech dla systemu automatycznego rozpoznawania mówcy (ang. Automatic Speaker Recognition – ASR). Etap generacji cech (parametryzacji sygnału mowy) jest fundamentalny w tego typu systemach, z uwagi na fakt, że unikatowy wektor cech ma decydujące znaczenie w procesie rozpoznawania. Zadaniem generatora cech jest opisanie sygnału mowy za pomocą możliwie mało licznego zbioru deskryptorów, bez utraty informacji istotnych z punktu widzenia rozpoznawania mówcy. Ponadto parametryzacja powinna wykazywać odporność na warunki akustyczne i techniczne rejestracji oraz na zawartość lingwistyczną rejestrowanego materiału. Badania przedstawione w referacie koncentrowały się przede wszystkim na wielokryterialnej optymalizacji wybranych parametrów generatora cech opartego na analizie cepstralnej, uwzgledniającej dodatkowo selekcję cech. Oceny otrzymanych wyników dokonano w oparciu o analizę składników głównych (ang. Principal Component Analysis – PCA) zbioru deskryptorów wyznaczonych dla próbek głosu pochodzących od 24 mówców.

9

Modelowanie i optymalizacja generatora cech dla systemu rozpoznawania mówcy

Majda E., Dobrowolski A. P., Smólski B. L.

Biuletyn Wojskowej Akademii Technicznej

|

2012

|

Vol. 61, nr 4

153-168

PL

W pracy przedstawiono zagadnienia związane z modelowaniem i optymalizacją generatora cech dla systemu automatycznego rozpoznawania mówcy (ang. Automatic Speaker Recognition - ASR). Etap generacji cech (parametryzacji sygnału mowy) jest fundamentalny w tego typu systemach, z uwagi na fakt, że unikatowy wektor cech ma decydujące znaczenie w procesie rozpoznawania. Zadaniem generatora cech jest opisanie sygnału mowy za pomocą możliwie mało licznego zbioru deskryptorów, bez utraty informacji istotnych z punktu widzenia rozpoznawania mówcy. Ponadto parametryzacja powinna wykazywać odporność na warunki akustyczne i techniczne rejestracji oraz na zawartość lingwistyczną rejestrowanego materiału. Badania przedstawione w referacie koncentrowały się przede wszystkim na wielokryterialnej optymalizacji wybranych parametrów generatora cech opartego na analizie cepstralnej, uwzględniającej dodatkowo selekcję cech. Oceny otrzymanych wyników dokonano w oparciu o analizę składników głównych (ang. Principal Component Analysis - PCA) zbioru deskryptorów wyznaczonych dla próbek głosu pochodzących od 24 mówców.

EN

The paper presents issues related to modeling and optimization of the features generator for the speaker recognition system (ASR - Automatic Speakers Recognition). Parameterization's stage of the speech signal (features generation) is fundamental in this type of systems, due to the fact that the unique vector of features is crucial in the process of recognition. The task is to describe the speech signal using descriptors as little as possible, without loss of relevant information to the speaker recognition. In addition, parametrization should have robust to acoustic and technical registration conditions and the recorded linguistic material. The research presented in this paper is focused primarily on the multicriteria optimization of selected parameters of the features generator based on cepstral analysis, additionally allowing features selection. Finally, evaluation of the results was based on the analysis of main components, a set of descriptors for the samples voice acquired from 24 speakers.

10

Exploratory data analysis methods for comparison of drug dissolution profiles

Enăchescu C.

Biocybernetics and Biomedical Engineering

|

2010

|

Vol. 30, no. 4

17-27

EN

A new approach for 'similarity' testing through comparison of drug products dissolution profiles, based on multivariate data analysis is presented. The dissolution curves corresponding to three products containing oxicams (piroxicam, meloxicam and tenoxicam) as oral solid dosage forms were obtained by dissolution tests at multiple pre-specified time points and in different compendial media. Dissolution data was simultaneously subjected to principal component and cluster analysis and comparisons between the dissolution characteristics of different products were carried out. All the results were compared with information provided by the difference (f1) and similarity (f2) factor tests. Unlike the f2 criterion, the proposed methods reflect variability within the individual dissolution curves, being also highly sensitive to profile variations.

11

Principal Component Analysis (PCA) for Feature Selection at the Diagnosis of Electrical Circuits

Sałat R., Osowski S., Siwek K.

Przegląd Elektrotechniczny

|

2003

|

R. 79, nr 10

667-670

EN

The paper presents the application of the Principal Component Analysis (PCA) to the analysis and comparison of different multidimensional feature vectors describing the given analogue circuit for the diagnosis purposes. The diagnosis is regarded here as the recognition of the faulty element in the circuit. Usually the PCA results reveal that only few first principal components are important from the point of view of the reconstruction of the compressed original information. These components contain normally more than 90% of the whole information. Presenting the investigated process on the plane composed by the first two components can reveal the ability of the features to discriminate between different states of the electrical circuit. On the basis of this we can asses the usefulness of these features for the diagnosis purposes.

PL

Praca przedstawia zastosowanie transformacji według składników głównych (PCA) w selekcji cech wzorca najlepiej charakteryzujących obwód analogowy podlegający diagnostyce. Diagnostyka jest rozumiana jako zadanie rozpoznania uszkodzonego elementu w obwodzie, gdzie pod pojęciem uszkodzenia rozumie się zmianę wartości parametru elementu poza przyjęty zakres tolerancji. Analiza PCA pozwala określić rozkład wartości głównych, z których zwykle dwa są dominujące. Rzutując dane pomiarowe na te dwa najważniejsze składniki główne otrzymuje się wizualizację rozkładu danych wielowymiarowych na płaszczyźnie. Z analizy rozkładów odpowiadających różnym wektorom cech wzorca można wnioskować o lepszym lub gorszym zróżnicowaniu cech diagnostycznych wybranej reprezentacji wzorca. Przedstawiona metoda została zilustrowana na przykładzie diagnostyki analogowego filtru RC drugiego rzędu.