Araştırma Çıktıları | WoS | Scopus | TR-Dizin | PubMed
Permanent URI for this communityhttps://hdl.handle.net/20.500.14719/1741
Browse
18 results
Search Results
Publication Metadata only Hyper-Spectral Image Segmentation Using Spectral Clustering With Covariance Descriptors(SPIE-INT SOC OPTICAL ENGINEERING, 2009) Kursun, Olcay; Karabiber, Fethullah; Koc, Cemalettin; Bal, Abdullah; Astola, JT; Egiazarian, KO; Nasrabadi, NM; Rizvi, SA; Bahcesehir University; Istanbul University; Gebze Technical University; Yildiz Technical UniversityImage segmentation is an important and difficult computer vision problem. Hyper-spectral images pose even more difficulty due to their high-dimensionality. Spectral clustering (SC) is a recently popular clustering/segmentation algorithm. In general, SC lifts the data to a high dimensional space, also known as the kernel trick, then derive eigenvectors in this new space, and finally using these new dimensions partition the data into clusters. We demonstrate that SC works efficiently when combined with covariance descriptors that can be used to assess pixelwise similarities rather than in the high-dimensional Euclidean space. We present the formulations and some preliminary results of the proposed hybrid image segmentation method for hyper-spectral images.Publication Metadata only Improving Automatic Emotion Recognition from Speech Signals(ISCA-INT SPEECH COMMUNICATION ASSOC, 2009) Bozkurt, Elif; Erzin, Engin; Erdem, Cigdem Eroglu; Erdem, A. Tanju; Koc University; Bahcesehir University; Ozyegin UniversityWe present a speech signal driven emotion recognition system. Our system is trained and tested with the INTERSPEECH 2009 Emotion Challenge corpus, which includes spontaneous and emotionally rich recordings. The challenge includes classifier and feature sub-challenges with five-class and two-class classification problems. We investigate prosody related, spectral and HMM-based features for the evaluation of emotion recognition with Gaussian mixture model (GMM) based classifiers. Spectral features consist of mel-scale cepstral coefficients (MFCC), line spectral frequency (LSF) features and their derivatives, whereas prosody-related features consist of mean normalized values of pitch, first derivative of pitch and intensity. Unsupervised training of HMM structures are employed to define prosody related temporal features for the emotion recognition problem. We also investigate data fusion of different features and decision fusion of different classifiers, which are not well studied for emotion recognition framework. Experimental results of automatic emotion recognition with the INTERSPEECH 2009 Emotion Challenge corpus are presented.Publication Metadata only An anomaly intrusion detection approach using Cellular Neural Networks(SPRINGER-VERLAG BERLIN, 2006) Yang, Zhongxue; Karahoca, Adem; Levi, A; Savas, E; Yenigun, H; Balcisory, S; Saygin, Y; Nanjing Xiaozhuang University; Bahcesehir UniversityThis paper presents an anomaly detection approach for the network intrusion detection based on Cellular Neural Networks (CNN) model. CNN has features with multi-dimensional array of neurons and local interconnections among cells. Recurrent Perceptron Learning Algorithm (RPLA) is used to learn the templates and bias in CNN classifier. Experiments with KDD Cup 1999 network traffic connections which have been preprocessed with methods of features selection and normalization have shown that CNN model is effective for intrusion detection. In contrast to back propagation neural network, CNN model exhibits an excellent performance owing to the higher attack detection rate with lower false positive rate.Publication Metadata only Network intrusion detection by using Cellular Neural Network with Tabu Search(IEEE COMPUTER SOC, 2008) Yang, Zhongxue; Karahoca, Adem; Yang, Ning; Aydin, Nizamettin; Stoica, A; Arslan, T; Howard, D; Higuchi, T; ElRayis, A; Nanjing Xiaozhuang University; Bahcesehir UniversityThis paper presents a novel Cellular Neural Network (CNN) templates learning approach based on Tabu Search (TS) for detecting network intrusions. The TS method was applied to CNN with symmetric templates and was verified by simulations. Simulation experiments on intrusion detection have shown that the TS-based template learning algorithm exhibits superior performance in computation time to find the optimal solution and in the solution quality in contrast to the algorithm of genetic algorithm (GA) and simulated annealing (SA).Publication Metadata only Review of Communication Systems for Ingestible Miniaturized Integrated Sensor Microsystems(IEEE COMPUTER SOC, 2009) Aydin, N.; Arslan, T.; Stoica, A; Arslan, T; Huntsberger, T; Botez, P; Erdogan, AT; ElRayis, AO; Bahcesehir University; University of EdinburghRecently, there is a growing interest for miniaturized telemetry systems for a variety of biomedical applications such as the development of wireless sensor systems that can be integrated into a noninvasive capsule format to perform endoscopic functions within the gastrointestinal tract. An important issue in such systems is the real time transmission of the data out of body wirelessly. Developments in system-on-chip and wireless technologies have led to complex electronic systems to be miniaturized to size of ingestible capsule and implantable microsystems. Inevitably such miniaturized complex systems impose some constraints on the case of an ingestible diagnostic capsule. It is desirable that system be wireless, programmable, and reusable. In this paper, we review some possible wireless interface methods and describe a programmable wireless direct sequence spread spectrum link developed for such an ingestible microsystem.Publication Metadata only A robust image watermarking based on time-frequency(IEEE, 2007) Ozturk, Mahmut; Akan, Aydin; Cekic, Yalcin; Istanbul University; Bahcesehir UniversityWatermarking techniques are proposed as a solution to copyright protection of digital media files. In this work, a new and robust watermarking method that is based on time-frequency (TF) representations is presented. We use the discrete evolutionary transform to represent an image in the TF domain. A watermark is embedded onto selected cells in the joint TF domain. Hence by combining the advantages of spatial and spectral domain watermarking methods, a robust and perceptual watermarking algorithm is presented.Publication Metadata only A threshold free clustering algorithm for robust unsupervised classification(IEEE COMPUTER SOC, 2007) Temel, Turgay; Aydin, Nizamettin; Stoica, A; Arslan, T; Howard, D; Kim, TH; ElRayis, A; Fatih University; Bahcesehir UniversityA new information-theoretic, unsupervised, subtractive clustering algorithm is proposed. The algorithm eliminates threshold constraint to detect possible cluster members. Cluster centers are formed with minimum entropy. Instead of using a fixed-threshold, a decision region is formed with the use of maximum mutual information. Cluster members are chosen with a relative-cost assigned in partitions of data set. The algorithm yields more reliably distributed cluster numbers in statistical sense, hence reducing further computation for validation, which is justified for a set of synthetic data.Publication Metadata only Fraud Detection Using an Adaptive Neuro-Fuzzy Inference System in Mobile Telecommunication Networks(OLD CITY PUBLISHING INC, 2009) Sanver, Mert; Karahoca, Adem; Stanford University; Bahcesehir UniversityGSM (Global Services of Mobile Communications) 1800 licenses were granted in the beginning of the 2000's in Turkey. Especially in the installation phase of the wireless telecom services, fraud usage can be an important source of revenue loss. Fraud can be defined as a dishonest or illegal use of services, with the intention to avoid service charges. Fraud detection is the name of the activities to identify unauthorized usage and prevent losses for the mobile network operators'. Mobile phone user's intentions may be predicted by the call detail records (CDRs) by using data mining (DM) techniques. This study compares various data mining techniques to obtain the best practical solution for the telecom fraud detection and offers the Adaptive Neuro Fuzzy Inference (ANFIS) method as a means to efficient fraud detection. In the test run, shown that ANFIS has provided sensitivity of 97% and specificity of 99%, where it classified 98.33% of the instances correctly.Publication Metadata only Prioritization of human capital measurement indicators using fuzzy AHP(PERGAMON-ELSEVIER SCIENCE LTD, 2007) Bozbura, F. Tunc; Beskese, Ahmet; Kahraman, Cengiz; Bahcesehir University; Istanbul Technical UniversityPeople in an organization constitute an important and essential asset which tremendously contributes to development and growth of that company by the help of their collective attitudes, skills and abilities. This is why the human capital (HC) can be considered the most important sub-dimension of the intellectual capital. Since you cannot manage what you cannot control, and you cannot control what you do not measure, the measurement of HC is a very important issue. This study aims at defining a methodology to improve the quality of prioritization of HC measurement indicators under fuzziness. To do so, a methodology based on the extent fuzzy analytic hierarchy process (AHP) is proposed. Within the model, five main attributes, talent, strategical integration, cultural relevance, knowledge management, and leadership, their sub-attributes, and 20 indicators are defined. The proposed model can be used for any country. However, the results obtained in the numerical example reflect the situation of HC in Turkey, since the experts are asked to make their evaluations considering the cultural characteristics of Turkey. The results of the study indicate that creating results by using knowledge, employees' skills index, sharing and reporting knowledge, and succession rate of training programs are the four most important measurement indicators for the HC in Turkey. (C) 2006 Elsevier Ltd. All rights reserved.Publication Metadata only An improved odor recognition system using learning vector quantization with a new discriminant analysis(ACAD SCIENCES CZECH REPUBLIC, INST COMPUTER SCIENCE, 2007) Temel, Turgay; Karlik, Bekir; Bahcesehir University; Fatih UniversityA new pre-processing algorithm for improved discrimination of odor samples is proposed. The pre-processed odor sample outputs from two sensors are input using a learning-vector quantization (LVQ) classifier as a means of odor recognition to be employed within electronic nose applications. The proposed algorithm brings out highly scattered classes while minimizing the within-class scatter of the samples given an odor class. LVQ is observed to operate robustly and reliably in terms of variation of parameters of interest, mainly a learning parameter. Due to the increased performance along with computational simplicity and robustness, the scheme is suitable to sample-by-sample identification of olfactory sensory data and can be easily adapted to hierarchical processing with other sensory data in real-time.