Saban Ozer

A novel fragile watermarking based on Particle Swarm Optimization

In this paper, a novel fragile watermarking scheme based on discrete cosine transform (DCT) using particle swarm optimization (PSO) algorithm is presented. Embedding watermarks in frequency domain can usually be achieved by modifying the least significant bits (LSBs) of the transformation coefficients. After embedding process is completed, a number of rounding errors appear due to conversion of real numbers into integers in the process of transformation of image from frequency domain to spatial domain. A population based stochastic optimization technique (PSO) is proposed to correct these rounding errors. Simulation results show the feasibility of employing PSO for watermarking and the accuracy of this novel method.

A novel clonal selection algorithm based fragile watermarking method

In this paper, a novel fragile watermarking method based on clonal selection algorithm (CSA), CLONALG, is presented. In Discrete Cosine Transform (DCT) based fragile watermarking techniques, there occurs some degree of rounding errors because of the conversion of real numbers into integers in the process of transformation of image from frequency domain to spatial domain. In this paper, the rounding errors caused by this transformation process are corrected by using CLONALG. Simulation results show that extracted watermark is obtained exactly the same as embedded watermark and optimum watermarked image transparency is achieved. In addition, the performance comparison of CLONALG and genetic algorithm (GA) based methods is realized.

System identification using Hammerstein model

In literature, various linear and nonlinear model structures are defined to identify the systems. Linear models such as Finite Impulse Response (FIR), Infinite Impulse Response (IIR) and Autoregressive (AR) are used in the situations that the input-output relation is signified through linear equivalence. However because of the nonlinear structure of the systems in real life, nonlinear models are developed. Volterra, Bilinear and polynomial autoregressive (PAR) are the examples of nonlinear models. In literature, there are also block oriented models to cascade the linear and nonlinear systems such as Hammerstein, Wiener and Hammerstein Wiener. These models are preferred because of practical use and effective prediction of wide nonlinear process. In this study, system identification applications of Hammerstein model that is cascade of nonlinear Volterra model and linear FIR model. Least mean Square (LMS) and Recursive Least Square (RLS) algorithms are used to identify the Hammerstein model parameters. Furthermore, The results are compared with the FIR model and Volterra model results to identify the success of Hammerstein model.

A comparison study of system identification using Hammerstein model

An attempt has been made in this paper to present performance analysis of a Hammerstein model for system identification area. Hammerstein model block structure is formed by cascade of linear and nonlinear parts. This study different from the studies in literature, focuses on the performance of Hammerstein block model that Second Order Volterra (SOV) Model is preferred instead of Memoryless Polynomial Nonlinear (MPN) as nonlinear part. In simulations, different systems are identified by proposed Hammerstein model which is optimized with classical and heuristic algorithms. Also, its performance is compared with different models.

A novel image watermarking method based on Discrete Cosine Transform using Genetic Algorithm

In this paper, a novel robust watermarking technique based on Discrete Cosine Transform using Genetic algorithm (GA) is presented. To obtain highest possible robustness without losing transparency, GA is used for adjustment of mid-band coefficients. Significant improvements in robustness under attacks and also in transparency are obtained from experimental results.

The Derivation of Theoretical Ultrasonic Doppler Blood Flow Spectrum

This paper deals with theoretical aspects of blood flow and the ultrasonic Doppler spectrum. Blood flow in the vessel can be considered as a pipe flow with a circular cross-section. Blood flow is, in general, complex and pulsatile. Navier-Stokes and the continuity equations are the governing equations for the blood flow, and the Naviers equations for the blood vessel. If the flow is pulsatile with a large Womersely number, the velocity profile becomes blunt due to the inertia forces. Since the ultrasonic Doppler flowmeter detects the velocities of blood cells across the vessel, Doppler power spectrum is dependent on the velocity profile. If the velocity profile is blunt, the Doppler power spectrum will be narrow banded. If the flow is sinusoidal, the Womersely number is only a parameter describing the pulsatility. Blood flow waveforms, however, differ from the sinusoidal wave. They consist of a steady flow component and many harmonics with amplitudes and phases. In this study, the influence of the harmonic contents and the Womersely number on velocity profiles and the Doppler power spectrum are examined.

An Improved Hammerstein Model for System Identification

System identification can easily model practical applications such as peer to peer (P2P) file-sharing traffic, driver assistance system, road traffic state, ethernet-based traffic flows. Therefore, system identification process can also be used in the smart city concept. This paper aims to improve Hammerstein model for system identification area. Hammerstein model block structure is formed by cascade of linear and nonlinear parts. Generally, MPN (Memoryless Polynomial Nonlinear) model for nonlinear part and FIR (Finite Impulse Response) or IIR (Infinite Impulse Response) model for linear part are preferred in Hammerstein models in literature. This study different from the studies in literature, focuses on the performance of Hammerstein block model that Second Order Volterra (SOV) Model is preferred instead of MPN as nonlinear part. In this context, a SOV based Hammerstein model structure is presented. In simulations, different nonlinear and linear systems are identified by our proposed Hammerstein model which is optimized with classical and heuristic algorithms. Also, its performance is compared with different linear and nonlinear models. We believe, we are the first ones to study and compare the performance between SOV based Hammerstein model and MPN based Hammerstein model. The main benefit of this study is that simulation results reveal the effectiveness and robustness of the proposed model. Therefore, Hammerstein model may be preferred to model different type of system in smart cities.

Identification of nonlinear systems using Clonal Selection algorithm

In this work, Clonal Selection algorithm (CSA) has been applied to adaptive identification of nonlinear systems and compared its performance to that of Genetic algorithm (GA). Nonlinear Box-Jenkins system which is frequently used as a benchmark example for testing in literature and a parametric nonlinear bilinear system have been identified using these algorithms. The simulation results have shown that nonlinear systems can be identified using CSA with low modeling error.

A new multi volterra system model based on artificial neural network for nonlinear system identification

In this paper, a new multi Volterra system model based on artificial neural network for nonlinear system identification is proposed. In this work, our attention is especially focused on behaviours of system which aren’t suffer from noise added to input and produce desired output. The proposed model were tested with Gaussian distribution noise signals at different levels and compared with single Volterra systems. The proposed method is illustrated by simulations. These simulations indicate that the performance of proposed model is better than single Volterra systems.

An artificial neural network based image sharpness criterion

Because sharply focused images inherently contain more information than defocused images, automatically obtaining the sharp image of a scene is an important task in computer vision. A camera can be sharply focused on an object point in different ways. To automate this task, a criterion function (CF) is needed to measure the sharpness of focus. Better performance can be achieved by computing more than one CFs and making judgements based on all of them rather than only one of them. This paper deals with the use of feedforward neural networks trained by LevenbergMarquardt algorithm to compute three different CFs for measuring the sharpness of images. The developed technique is employed on different images to calculate the sharpness of them and its performance is discussed.