Suhap Şahin

Neural network training based on FPGA with floating point number format and it’s performance

In this paper, two-layered feed forward artificial neural network’s (ANN) training by back propagation and its implementation on FPGA (field programmable gate array) using floating point number format with different bit lengths are remarked based on EX-OR problem. In the study, being suitable with the parallel data-processing specification on ANN’s nature, it is especially ensured to realize ANN training operations parallel over FPGA. On the training, Virtex2vp30 chip of Xilinx FPGA family is used. The network created on FPGA is coded by using VHDL. By comparing the results to available literature, the technique developed here proved to consume less space for the subjected ANN training which has the same structure and bit length, it is shown to have better performance.

RoadRunneR: A Small and Fast Bitslice Block Cipher for Low Cost 8-Bit Processors

Designing block ciphers targeting resource constrained 8-bit CPUs is a challenging problem. There are many recent lightweight ciphers designed for better performance in hardware. On the other hand, most software efficient lightweight ciphers either lack a security proof or have a low security margin. To fill the gap, we present RoadRunneR which is an efficient block cipher in 8-bit software, and its security is provable against differential and linear attacks. RoadRunneR has lowest code size in Atmels ATtiny45, except NSAs design SPECK, which has no security proof. Moreover, we propose a new metric for the fair comparison of block ciphers. This metric, called ST/A, is the first metric to use key length as a parameter to rank ciphers of different key length in a fair way. By using ST/A and other metrics in the literature, we show that RoadRunneR is competitive among existing ciphers on ATtiny45.

Edge detection and reduction of brightness of students’ bubble form images

Optical Mark Recognition (OMR) is a traditional data input technique and an important human computer interaction technique which is widely used in education testing. This paper proposes a new idea for grading multiple-choice test which is based on a camera on smartphone. The system key techniques and relevant implementations, which include the image scan, edge detection and reduction of brightness on colorful bubble form images, are presented.

A Method for Localization of Computational Node and Proxy Server in Educational Data Synchronization

Localization methods enable location estimation accurately and provide location information about mobile devices, people, cars, data and equipment. Accurate location detection is a vital process for most of location-based applications such as emergency rescue, in-building guidance, security services, and product tracking in hospitals. This paper addresses localization of student/teacher tablets and school level proxy servers for educational data synchronization. For this purpose, locations of proxy servers and tablets are detected using Android Location API. After localization, if tablets are outside of school, they could access cloud server directly to get educational data and if tablets are in school, they could access data via proxy server. Experimental results show that the proposed technique increases the efficiency in data transfers between the end users and cloud servers.

Computer Aided Wound Area Detection System for Dermatological Images

Research shows that in the last decade, the focus on computer-assisted diagnoses on the skin disorders has increased significantly as a result of the improvements in skin imaging technology and the development of compatible image processing techniques. More accurate treatments provided by means of computer-assisted diagnostic systems increase the patients’ chances of recovery and survival. Image processing techniques used in these systems facilitate the detection of wound areas. In this study, a wound detection system using adaptive weighted median filter (AWMF), Otsu’s thresholding, and an implementation of the Canny edge detection algorithm using the Sobel kernel, respectively, is proposed for the detection of wound areas on dermatological images. The effectiveness of the system is tested on different dermatological datasets. Obtained values are analyzed with Peak Signal to Nose Ratio (PSNR) and Correlation Coefficient (CC) metrics and it was confirmed that the system works accurately on various datasets.

gDsDK*0 and gBsDK*0 coupling constants in QCD sum rules

In the present study, we calculate the strong coupling constants gDsDK*0(800) and gBsDK*0(800) within the three-point QCD sum rules approach. We evaluate the correlation function of the considered vertices taking into account both D[B] and K*0(800) mesons as off-shell states.