Hasan Demir

Dr. Hasan Demir


Research Focus & Metrics

Academic Career Length

9Years (2018-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
1 January 2018 in Electrica
Verified
Journal Article
Closed access
PDF file
HDHasan Demir
Hasan Demir
Herein, using support vector machines, texture images were classified based on the histogram of oriented gradients, from which feature vectors were obtained. In addition, the success rate was examined for the feature vectors with different dimensions and the minimum length of a feature vector for performing classification was determined to be 288 elements.

1 May 2018 in Tehnicki vjesnik - Technical Gazette
Verified
Journal Article
Open access
PDF file
ÖAÖMER AKGÜN
ÖMER AKGÜN
Aydin AkanAydin Akan
Aydin Akan
HDHasan Demir
Hasan Demir
TATahir Cetin AKINCI
Tahir Cetin AKINCI
In this study, a gait device was used for gathering data. A group comprising control group and ALS patients was requested to walk using this device. Gait signals of the control group individuals and ALS patients taken from their left...

31 July 2021 in European Journal of Engineering and Applied Sciences
Verified
Journal Article
Closed access
PDF file
FAFatih Alteki̇n
Fatih Alteki̇n
HDHasan Demir
Hasan Demir
In this article, image processing techniques to detect facial emotion expressions are examined. Studies done to detect facial expression are given in stages. The success of the convolutional neural networks (CNN) method in emotional expression has been investigated. A set of 981 CK + pictures containing human faces in 7 emotion categories was used. The success rates when using HOG, LBP, Wavelet feature of images and the original state of the images in the data set were compared.

31 December 2024 in European Journal of Engineering and Applied Sciences
Verified
Journal Article
Open access
PDF file
HAHafiz Alisoy
Hafiz Alisoy
HDHasan Demir
Hasan Demir
GAGülizar Alisoy
Gülizar Alisoy
The development of innovative electrojet technologies based on the formation of sequential flows of monodisperse droplets in the working environment remains among the current problems today. In this study, it has been determined that in the “needle-plane” electrode system where corona discharge occurs. The charging kinetics of spherical particles are investigated in this system. These particles were created under equal conditions for various dielectric liquids. The spherical particles had different radii and different relative dielectric permittivities. The study determined that the charging kinetics of these particles differ. This difference depends on the polarity of the small radius electrode. In particular, the effective ionization coefficient in an electronegative gas environment (e.g. in air) where corona discharge occurs, it has been shown that when the needle electrode has a negative polarity, the amount of charge accumulated on the spherical particle is higher compared to the positive polarity case.

14 March 2025 in Journal of Balkan Science and Technology
Verified
Journal Article
Closed access
HAHafiz Ali̇sOy
Hafiz Ali̇sOy
HDHasan Demir
Hasan Demir
SAShahriyar Ahmedov1
Shahriyar Ahmedov1
This study creates the design of a 10th order Butterworth bandpass filter based on an operational amplifier. It also provides the analysis of the results obtained in this context. Higher order filters are constructed using a combination of second and third order filters. Four parameters need to be specified in order to design a Butterworth bandpass filter, namely Ap (passband attenuation in dB), As (stopband attenuation in dB), fp (frequency at which Ap occurs), and fs (frequency at which As occurs). The design procedure involves two steps: one is to find the required filter order, and the other is to find the scaling factor which is to be applied to the formatted parameter values. The Butterworth bandpass filter is a combination of a low-pass filter and a high-pass filter. For a Butterworth low-pass filter, the resistance is approximately of 100 kΩ and the capacitor value is found by inverse scaling with the frequency and the chosen resistance value. For a Butterworth high pass filter, the used capacitor value is approximately of 0.05 µF and the resistor value is found by inverse scaling with the frequency and the chosen resistance value.

20 November 2026 in IJACI : International Journal of Advanced Computing and Informatics
Verified
Journal Article
Open access
AKAbdulfatah Kalaje
Abdulfatah Kalaje
HDHasan Demir
Hasan Demir
Alzheimer disease (AD) is a neurodegenerative disorder that affects memory, cognition, and behavior, representing that leads to dementia with older adults. AD detection is performed using various deep learning models applied to magnetic resonance images (MRI). The feature vector is extracted by using 2D discrete wavelet transformation (DWT) to capture approximation, horizontal, vertical, and diagonal coefficients of each image. Then 1D, 2D, and 3D deep learning models are employed, and their performance is evaluated and compared using evaluation metrics, including accuracy, precision, sensitivity, specificity, and F1-score. The 1D model included 1D convolutional neural network (CNN), while 2D CNN, CNN combined with long short-term memory (CNN+LSTM), ResNet-18, ResNet-34, and MobileNetV1 architectures are also implemented. 3D CNN and 3D CNN+LSTM are used for 3D models. Experimental results showed that the 3D-CNN model outperforms all other architectures because it uses the approximation horizontal, vertical, and diagonal components of 2D DWT as one input block, achieving an accuracy of 99.88%, precision of 99.66%, sensitivity of 99.89%, specificity of 99.86%, and an F1-score of 99.76%.