Wyniki wyszukiwania - BazTech

Ograniczanie wyników

Znaleziono wyników: 3

Liczba wyników na stronie

Wyniki wyszukiwania

Sortuj według:

Ogranicz wyniki do:

Novel multiple pooling and local phase quantization stable feature extraction techniques for automated classification of brain infarcts

Dogan Sengul, Barua Prabal Datta, Baygin Mehmet, Chakraborty Subrata, Ciaccio Edward J., Tuncer Turker, Abd Kadir Khairul Azmi, Md Shah Mohammad Nazri, Azman Raja Rizal, Lee Chin Chew, Ng Kwan Hoong, Acharya U. Rajendra

Biocybernetics and Biomedical Engineering

2022

Vol. 42, no. 3

815--828

This study aims to introduce a hand-crafted machine learning method to classify ischemic and hemorrhagic strokes with satisfactory performance. In the first step of this work, a new CT brain for images dataset was collected for stroke patients. A highly accurate hand-crafted machine learning method is developed and tested for these cases. This model uses preprocessing, feature creation using a novel pooling method (it is named P9), a local phase quantization (LPQ) operator, and a Chi2-based selector responsible for selecting the most significant features. After that, classification is done using the k-nearest neighbor (kNN) classifier with ten-fold cross-validation (CV). The novel aspect of this model is the P9 pooling method. The inspiration for this pooling method was drawn from the deep learning models, where features are extracted with multiple layers using a convolution operator applied to the pooling method. However, pooling decompositions have a routing problem. The P9 pooling function creates nine decomposed models, hence the name. The LPQ feature extractor is applied to images to generate sub-bands for feature generation. The Chi2 selector is then employed to select the most significant features from the created feature vector, and these features are utilized for the classification using the k-nearest neighbor algorithm (kNN). The introduced P9-LPQ feature extraction-based learning model attained over 98% classification accuracy in all cases. The results obtained in this paper show that the proposed method can successfully classify stroke types. For this reason, the developed model can pre-diagnose stroke types in the future.

Automated detection of Parkinson's disease using minimum average maximum tree and singular value decomposition method with vowels

Tuncer Turker, Dogan Sengul, Acharya Udyavara Rajendra

Biocybernetics and Biomedical Engineering

2020

Vol. 40, no. 1

211--220

In this study, a novel method to automatically detect Parkinson's disease (PD) using vowels is proposed. A combination of minimum average maximum (MAMa) tree and singular value decomposition (SVD) are used to extract the salient features from the voice signals. A novel feature signal is constructed from 3 levels of MAMa tree in the preprocessing phase. The SVD operator is applied to the constructed signal for feature extraction. Then 50 most distinctive features are selected using relief feature selection technique. Finally, k nearest neighborhood (KNN) with 10-fold cross validation is used for the classification. We have achieved the highest classification accuracy rate of 92.46% using vowels with KNN classifier. The dataset used consists of 3 vowels for each person. To obtain individual results, post processing step is performed and best result of 96.83% is obtained with KNN classifier. The proposed method is ready to be tested with huge database and can aid the neurologists in the diagnosis of PD using vowels.

Improved reference image encryption methods based on 2k correction in the integer wavelet domain

Tuncer Turker, Dogan Sengul, Tadeusiewicz Ryszard, Pławiak Paweł

International Journal of Applied Mathematics and Computer Science

2019

Vol. 29, no. 4

817--829

Many visually meaningful image encryption (VMIE) methods have been proposed in the literature using reference encryption. However, the most important problems of these methods are low visual quality and blindness. Owing to the low visual quality, the pre-encrypted image can be analyzed simply from the reference image and, in order to decrypt nonblind methods, users should use original reference images. In this paper, two novel reference image encryption methods based on the integer DWT (discrete wavelet transform) using 2k correction are proposed. These methods are blind and have high visual quality, as well as short execution times. The main aim of the proposed methods is to solve the problem of the three VMIE methods existing in the literature. The proposed methods mainly consist of the integer DWT, pre-encrypted image embedding by kLSBs (k least significant bits) and 2k correction. In the decryption phase, the integer DWT and preencrypted image extraction with the mod operator are used. Peak signal-to-noise ratio (PSNR) measures the performances of the proposed methods. Experimental results clearly illustrate that the proposed methods improve the visual quality of the reference image encryption methods. Overall, 2k correction and kLSBs provide high visual quality and blindness.