Ulus Çevik

A Review on Feature Extraction for Speaker Recognition under Degraded Conditions

ABSTRACT Speech is a signal that includes speakers emotion, characteristic specification, phoneme-information etc. Various methods have been proposed for speaker recognition by extracting specifications of a given utterance. Among them, short-term cepstral features are used excessively in speech, and speaker recognition areas because of their low complexity, and high performance in controlled environments. On the other hand, their performances decrease dramatically under degraded conditions such as channel mismatch, additive noise, emotional variability, etc. In this paper, a literature review on speaker-specific information extraction from speech is presented by considering the latest studies offering solutions to the aforementioned problem. The studies are categorized in three groups considering their robustness against channel mismatch, additive noise, and other degradations such as vocal effort, emotion mismatch, etc. For a more understandable representation, they are also classified into two tables by utilizing their classification methods, and used data-sets.

Endoscopic instrument tracking for surgical simulation training in a controlled environment via a camera and a planar mirror

Minimally Invasive Surgery (MIS) has many advantages over traditional procedures and thus training with MIS tools via computer simulations has received much attention. These tools are generally grouped into two major categories: Physical training-boxes, and Computer vision/Virtual Reality (VR) tools. In this study, a computer vision based simulator is proposed which uses a training box that is composed of a single camera and a planar mirror. Occlusions are appropriately handled by the use of the epipoint geometry. The average 3D positional error was 0.96mm (±0.44mm) at 1280×960 resolution, and 1.18mm (±0.52mm) at 320×240. So, the error is minimally affected as the resolution decreases. The proposed method has some advantages over relevant literature methods, such as an improved accuracy (approximately 60%) even at low resolutions with a low processing time (approximately 30%). Therefore, the proposed method appears as a promising and low cost (approximately 50%) alternative for computer vision based MIS training tools.

An Automated Perspective-Projected View Module for Pixel-Based CSG Renderers

Abstract The generation of 3D solid objects, and more generally solid geometric modelling, is very important in Computer Aided Design (CAD). This paper presents a simple but effective algorithm for automated display of perspective views of Constructive Solid Geometry (CSG) scene models. This algorithm can be implemented as a module in such a way that it is easily integrated, without any modification, to the present systems of “pixel-based Z-Buffer” CSG renderers. An implementation of the algorithm for such a system is also given in the paper.

Selective surface normal estimation for volume rendering

AbstractThree-dimensional volumetric data is discrete by nature. It does not preserve continuous surface. Therefore, to render surfaces realistically, the normal vectors must be predicted before processing the image. During the rendering, if the normals are averaged directly, then the sharp edges are smoothed out as well. In this paper, we show how to avoid over-smoothing problems by analyzing a few neighbor voxels, and selectively averaging the normals to keep details, within a predefined kernel.

Overview of Computer Graphics and algorithms

Computer Graphics (CG) is the art of rendering, and visualizing images on the computer screens. In three-dimensional (3D) CG, a scene is first modeled geometrically, typically using triangles, and the computer is then used to calculate what the scene will look like from a specific view point at a particular instant. In CG, one of the major goals is to create photo-realistic images in real time. In recent years, Volume Visualization (VV) has attracted the attention of many researchers. VV techniques have been used to analyze and render 3D datasets, obtained from a variety of sources including medical scanners, and results of simulation of physical and synthetic phenomena, on the computer screen. Volume Graphics (VG) has proven itself as an independent graphics technology. A common purpose of VG is to achieve photo realistic rendering. To achieve this, reflections, shadows, refraction and perspective projections are all necessary elements since they occur naturally in the natural environment.

3D pose estimation for laparoscopic instruments: A single camera and a planar mirror approach

Minimally Invasive Surgery (MIS) with Laparoscopic instruments (LI) has many advantages over traditional procedures, and thus training of laparoscopic instruments via computer simulations has become important. In this study, a method is proposed for 3d pose estimation. For the proposed scheme, a computer vision based training simulator which used a training box composed of a single camera and a planar mirror was developed. This method has advantages over similar literature methods in terms of processing speed and capability of working even with low resolution images. Furthermore, the proposed approach can handle tool occlusions by using the epiline geometry.

Microcontroller-based Implementation of a Perspective View Module for Pixel-based CSG Renderers

This paper presents the implementation of a microcontroller-based perspective view module to be integrated, without any modification, to the pixel-based Z-buffer constructive solid geometry (CSG) renderers.