Wyniki wyszukiwania - BazTech

1

Finding robust transfer features for unsupervised domain adaptation

Gao Depeng, Wu Rui, Liu Jiafeng, Fan Xiaopeng, Tang Xianglong

International Journal of Applied Mathematics and Computer Science

|

2020

|

Vol. 30, no. 1

99--112

EN

An insufficient number or lack of training samples is a bottleneck in traditional machine learning and object recognition. Recently, unsupervised domain adaptation has been proposed and then widely applied for cross-domain object recognition, which can utilize the labeled samples from a source domain to improve the classification performance in a target domain where no labeled sample is available. The two domains have the same feature and label spaces but different distributions. Most existing approaches aim to learn new representations of samples in source and target domains by reducing the distribution discrepancy between domains while maximizing the covariance of all samples. However, they ignore subspace discrimination, which is essential for classification. Recently, some approaches have incorporated discriminative information of source samples, but the learned space tends to be overfitted on these samples, because they do not consider the structure information of target samples. Therefore, we propose a feature reduction approach to learn robust transfer features for reducing the distribution discrepancy between domains and preserving discriminative information of the source domain and the local structure of the target domain. Experimental results on several well-known cross-domain datasets show that the proposed method outperforms state-of-the-art techniques in most cases.

2

Developing an Intelligent Model for the Construction a Hip Shape Recognition System Based on 3D Body Measurement

Jin J.-F., Yang Y.-C., Zou F.-Y.

Fibres & Textiles in Eastern Europe

|

2016

|

Vol. 24, no. 5 (119)

110--118

EN

The purpose of this paper was to develop an intelligent recognition system consisting of a feature reduction method combining cluster and correlation analyses, and a probabilistic neural network (PNN) classifier to identify different types of hip shape from 3D measurement for each person. Firstly 28 items reflecting lower body part information of 300 female university students aging from 20 to 24 years were selected. The feature reduction method was employed to extract typical indices. Secondly hip shapes were subdivided into five types by a K-means cluster and analysis of variance (ANOVA). Finally the PNN was then trained to serve as a classifier for identifying five different hip shape types. The average classification accuracy of the scheme proposed was 97.37%, and its effectiveness was successfully validated by comparing with the BP and Support Vector Machine (SVM) scheme. Thus an intelligent recognition system was developed to make hip shape type classification of high-precision and time saving.

PL

Model łączy analizę skupień i korelacji oraz probabilistyczną sztuczną sieć neuronową dla identyfikacji różnych typów kształtów bioder opartą o pomiary 3D poszczególnych osób. Wyselekcjonowano 28 przypadków odzwierciedlających dolną część sylwetki 300 studentek w wieku od 20 do 24 lat. Zastosowano metodę redukcji poszczególnych właściwości dla wybrania typowych wskaźników. Następnie kształt bioder podzielono na 5 typów za pomocą algorytmu klastrowego i systemu ANOVA (analiza wariancji). Następnie przeprowadzono trening sieci neuronowej aby mogła posłużyć jako klasyfikator identyfikacji 5 różnych kształtów bioder. Przeciętna dokładność klasyfikacji proponowanego systemu wynosiła 97,37%, a efektywność była sukcesywnie sprawdzana przez porównanie schematów BP i SVM. W ten sposób stworzono inteligentny system rozpoznania typu kształtu bioder o dużej precyzji, pozwalający na oszczędność czasu.

3

Application of clustering techniques for visually evoked potentials based detection of vision impairments

Vijean V., M H., Yaacob S., Sulaiman M. N. B.

Biocybernetics and Biomedical Engineering

|

2014

|

Vol. 34, no. 3

169--177

EN

Visually evoked potentials (VEP) are evoked responses of the brain corresponding to a specific visual stimulus. Ophthalmologists often refer their patients to VEP test if the latter suffers any vision abnormalities that cannot be diagnosed using conventional analysis. By investigating the VEP responses, medical experts can narrow down the possible cause of the defect. Although this method provides valuable information to the medical practitioner, there are several drawbacks of the analysis that can affect the diagnosis result. The conventional averaging of the signals results in inter-trial variation between the VEP responses to be lost. This method also requires large number of trials, which causes fatigue in patients and reduces the diagnostic accuracy. Therefore, we have proposed a new method of analysis using statistical features derived from time and spectral space for the discrimination of vision impairments. Feature enhancement methods such as feature weighting and dimensional reduction are used to enhance the statistical features prior to the analysis. Four clustering methods are employed to increase the interclass separability of the control and myopic features while reducing the within class variability. The dimension of the weighted features is reduced using a combination of principal component analysis (PCA) and independent component analysis (ICA) techniques prior to classification. The proposed method is able to achieve 100% accuracy using extreme learning machine (ELM) and multi layer neural network (MLNN) classifiers.

4

Unsupervised Tracking, Roughness and Quantitative Indices

Pal S.K., Chakraborty D.

Fundamenta Informaticae

|

2013

|

Vol. 124, nr 1/2

63--90

EN

This paper presents a novel methodology for tracking a single moving object in a video sequence applying the concept of rough set theory. The novelty of this technique is that it does not consider any prior information about the video sequence unlike many existing techniques. The first target model is constructed using the median filtering based foreground detection technique and after that the target is reconstructed in every frame according to the rough set based feature reduction concept incorporating a measure of indiscernibility instead of indiscernibility matrix. The area of interest is initially defined roughly in every frame based on the object shift in the previous frames, and after reduction of redundant features the object is tracked. The measure of indiscernibility of a feature is defined based on its degree of belonging (DoB) to the target. Three quantitative indices based on rough sets, feature similarity and Bhattacharya distance are proposed to evaluate the performance of tracking and detect the mis-tracked frames in the process of tracking to make those corrected. Unlike many existing measures, the proposed ones do not require to know the ground truth or trajectory of the video sequence. Extensive experimental results are given to demonstrate the effectiveness of the method. Comparative performance is demonstrated both visually and quantitatively.

5

Linear discriminant analysis as EEG features reduction technique for brain-computer interfaces

Rak R. J., Majkowski A., Kołodziej M.

Przegląd Elektrotechniczny

|

2012

|

R. 88, nr 3a

28-30

EN

BCI systems analyze the EEG signal and translate patient intentions into simple commands. Signal processing methods are very important in such systems. Signal processing covers: preprocessing, feature extraction, feature selection and classification. In the article authors present the results of implementing linear discriminant analysis as a feature reduction technique for BCI systems.

PL

Systemy BCI analizują sygnał EEG i tłumaczą intencje użytkownika na proste polecenia. Ważnym elementem systemów BCI jest przetwarzanie sygnału. Obejmuje ono: przetwarzanie wstępne, ekstrakcję cech, selekcję cech i klasyfikację. W artykule autorzy prezentują wyniki badań z zastosowaniem liniowej analizy dyskryminacyjnej jako narzędzia do redukcji cech.