Search

In this study, the Internet of things is defined and lightweight protocols that the devices can use to communicate with each other are specified, and message queue telemetry transport (MQTT) protocol, which is designed to be used for restricted devices, is detailed. In the scope of this paper, the studies about the authentication of the objects and the confidentiality of the data have been examined, some suggestions have been made against the identified security vulnerabilities. In the scope of this paper, as the test environment, Mosquitto, the server of the MQTT protocol, and MongoDB in order to store the data were installed. In the prepared test environment, the objects were communicated with each other via the MQTT protocol and identity verification and security vulnerabilities were tried to be detected.

Date of Conference: 16-19 May 2016 Conference Name: IEEE 24th Signal Processing and Communications Applications Conference, SIU 2016 In this paper, we study multi-class classification of tweets, where we introduce highly efficient dimensionality reduction techniques suitable for online processing of high dimensional feature vectors generated from freely-worded text. As for the real life case study, we work on tweets in the Turkish language, however, our methods are generic and can be used for other languages as clearly explained in the paper. Since we work on a real life application and the tweets are freely worded, we introduce text correction, normalization and root finding algorithms. Although text processing and classification are highly important due to many applications such as emotion recognition, advertisement selection, etc., online classification and regression algorithms over text are limited due to need for high dimensional vectors to represent natural text inputs. We overcome such limitations by showing that randomized projections and piecewise linear models can be efficiently leveraged to significantly reduce the computational cost for feature vector extraction from the tweets. Hence, we can perform multi-class tweet classification and regression in real time. We demonstrate our results over tweets collected from a real life case study where the tweets are freely-worded, e.g., with emoticons, shortened words, special characters, etc., and are unstructured. We implement several well-known machine learning algorithms as well as novel regression methods and demonstrate that we can significantly reduce the computational complexity with insignificant change in the classification and regression performance.

Date of Conference: 2-5 May 2018 In this paper, we study online nonlinear data regression and propose a highly efficient long short term memory (LSTM) network based architecture. Here, we also introduce on-line training algorithms to learn the parameters of the introduced architecture. We first propose an LSTM based architecture for data regression. To diminish the complexity of this architecture, we use an energy efficient operator (ef-operator) instead of the multiplication operation. We then factorize the matrices of the LSTM network to reduce the total number of parameters to be learned. In order to train the parameters of this structure, we introduce online learning methods based on the exponentiated gradient (EG) and stochastic gradient descent (SGD) algorithms. Experimental results demonstrate considerable performance and efficiency improvements provided by the introduced architecture.

Date of Conference: 2-5 May 2018 In this paper, we investigate online parameter learning for Long Short Term Memory (LSTM) architectures in distributed networks. Here, we first introduce an LSTM based structure for regression. Then, we provide the equations of this structure in a state space form for each node in our network. Using this form, we then learn the parameters via our Distributed Particle Filtering based (DPF) training method. Our training method asymptotically converges to the optimal parameter set provided that we satisfy certain trivial requirements. While achieving this performance, our training method only causes a computational load that is similar to the efficient first order gradient based training methods. Through real life experiments, we show substantial performance gains compared to the conventional methods.

Synchrohasor measurement is a common tool to monitor system stability and detect faults in electric power systems and various Phasor Measurement Units (PMUs) have been developed for this purpose. Parameter computation methods of PMUs specifically designed for rapidly time-varying voltage cases caused by nonlinear dynamic loads have been an attractive research topic. In this paper, instantaneous amplitudes and phases of the voltage signals are estimated using a method based on quadrature amplitude modulation complying with M class accuracy criteria defined by the IEEE Standard C37.118. It has been shown on simulated voltage data that the proposed method provides synchrophasor results satisfying the accuracy limits defined in IEEE Standard C37.118. The results of the proposed method have also been compared with those of the DFT method and the proposed method has been shown to be more advantages in terms of satisfying the accuracy limits and reducing processing time.

In this study, a method is proposed to determine the parameters of Proportional Integral Derivative Acceleration (PIDA) controller, which are used in the control of higher degree systems, via Particle Swarm Optimization (PSO) algorithm. Classic controller methods may sometimes be insufficient, especially in the control of higher degree systems. In PIDA controller design, heuristic optimization algorithms, which is one of the recommended methods for these systems, provides successful results. The performance of the PIDA controller obtained by the proposed method in the paper is compared with the gradient search and genetic algorithms in the literature. The unit step response and the frequency response of the controller show the effectiveness of the method.

The multiuser detection (MUD) at the receiver end of the Space Division Multiple Access aided Orthogonal Frequency Division Multiplexing (SDMA-OFDM) systems is very crucial for receiver design technologies. Due to the detailed search mechanism, the maximum likelihood (ML) MUD that provides optimal performance is limited to a dramatically increased computational complexity. The QR decomposition with M-algorithm maximum likelihood detection (QRM-MLD) MUD reduces the computational complexity compare with the computational complexity of ML MUD. The minimum mean squared error (MMSE) MUD with low complexity exhibit poor performance. MMSE-successive interference cancellation (SIC) MUD with almost the same complexity provides better bit error rate (BER) performance than linear MMSE MUD. In this paper, Hybrid QRM-MLD-MMSE with SIC adaptive multiuser detection has been proposed for signal detection. The simulation results show that the proposed method significantly reduces the complexity of the system and controls the BER performance of the system at the desired level.

Pedestrian Detection (PD) is one the most studied application areas of object detection. Although an important effort is already given to develop advanced classifiers for self-driving cars (SDC) especially in recent years, studies about onboard railway systems remain very limited. Although the similarity of railway and highway scenarios suggests that systems developed for SDC-PD can be used for rail vehicles, it is also necessary to examine the low level of generalization ability and the high degree of dependence of the PD test performance on the training set. In this paper, the feasibility of PD systems, whose effectiveness on SDC is proved, was examined on railway systems. For this purpose, SDC-PD systems, created with different training benchmarks such as Caltech and INRIA, have been applied to railway system scenes with various difficulties and the results have been examined. The findings show that the direct implementation of classifiers for SDC does not provide sufficient success to come up with specific challenges to railway systems.

In traditional networking architecture, for each problem a specific protocol is dedicated to network devices and these devices are configured manually. Since they are hardware-based, the implementation of new ideas and technologies to these devices is very difficult. However, today's increasing use of smartphones and other mobile data devices has increased the data traffic, and with the addition of service diversity and the needs of users, the existing network structure has become much more complex. Software Defined Networks (SDN) is a new network paradigm that allows a programmable central controller mechanism which separates the control plane from the data plane. With the advantage of a centralized control mechanism, SDN ensures that network structures can respond more quickly to current demands, be more flexible and reprogrammable. In addition, all these processes are implemented at a lower cost compared to traditional networking architecture. To evaluate the performance of SDN architectures, testbeds are used by researchers. In this paper, a small-scale and cost-effective SDN testbed is designed and implemented. The testing environment is realized using Raspberry Pi 3, Floodlight and Open vSwitch. Controller performance has been observed with the MACtracker application.

Mobile devices begin to spread increasingly recently to offer a lot of services which personal computers offer. This condition has led to increase in the number of security threats in mobile devices and services. In this paper, it has been made a research on mobile malware and malware detection techniques. Within the scope of the study, a a permission based detection system based on the machine learning methods for Android malware was developed. Developed system is analyzed by using Random Forest, Support Vector Machine and Artificial Neural Networks algorithms.

Nowadays, along with the development of information technologies, storage and analysis of biomedical datasets are easy in health sector. In this area, Machine Learning methods provide a great contribution for evaluation and interpretation of data. In this paper, in addition to Support Vector Machines, Decision Tree, K-Nearest Neighbors, Naive Bayes and Dictionary Learning methods, Random Feature Subspaces (RDL) and Random Instance Subspaces (BDL) methods which are the ensembles of Dictionary Learning are used in biomedical data classification. In the test results, SVM and Dictionary Learning methods, RDL and BDL, which are generated using random feature/instance subspaces achieve optimum accuracy results.

25th Signal Processing and Communications Applications Conference (SIU) -- MAY 15-18, 2017 -- Antalya, TURKEY WOS: 000413813100519 Along with the widespread use of technology, there have been many studies on human-computer relation recently. Study of the effects of computer systems on human education can be shown as an example of this. In this paper, a reliable and real time tangible user interface system developed for the air drum game which is used both for game and computer based training programs is presented. A user can perform a play action in six different directions using electronic drum sticks. A developed algorithm detects the direction of moving stick by processing sensor data taken from these sticks. Information about the stick which is moved and its direction are sent to the sound system to play the sound file of the relevant drum instrument. In this way the user is presented with a realistic drum experience. The developed system has been tested by different users in real world and its performance has been reported. The results verify the reliability and usability of the electronic drum game. Turk Telekom, Arcelik A S, Aselsan, ARGENIT, HAVELSAN, NETAS, Adresgezgini, IEEE Turkey Sect, AVCR Informat Technologies, Cisco, i2i Syst, Integrated Syst & Syst Design, ENOVAS, FiGES Engn, MS Spektral, Istanbul Teknik Univ

25th Signal Processing and Communications Applications Conference (SIU) -- MAY 15-18, 2017 -- Antalya, TURKEY, WOS: 000413813100592, In this paper, a study which automatically detects four open string notes of the violin from the recorded violin sounds is presented for the purpose of using in the beginner level violin education. First, in order to be able to analyze the violin sound recordings, it is necessary for the music piece to be segmented into the notes forming the music piece itself For this, it is necessary to correctly determine the starting times (onsets) of the notes. Because of the envelope characteristic that the violin sound has, it is more difficult to detect the note onset than other musical instrument sounds such as piano and guitar. By using the complex spectral difference method, the note onsets were determined, then by determining the fundamental frequency the note of the violin sound was found. The proposed system was applied to the open string sound tracks obtained from eight people by using the iPad-based professional recording system, and the results showed that the proposed system worked successfully., Turk Telekom, Arcelik A S, Aselsan, ARGENIT, HAVELSAN, NETAS, Adresgezgini, IEEE Turkey Sect, AVCR Informat Technologies, Cisco, i2i Syst, Integrated Syst & Syst Design, ENOVAS, FiGES Engn, MS Spektral, Istanbul Teknik Univ

Thanks to the new technologies of remote sensors, remote sensor data which increases in size exponentially and transforms to big data with increasing spatial and spectral resolutions causes great difficulties in storage, transportation and processing. It has become a necessity to implement parallel approaches instead of traditional methods, which are inadequate in critical applications when real/near real-time analysis of relevant large data is needed. In this paper, high performance computing approaches such as Multi-core, FPGA, GPU, Cluster and Cloud Computing are investigated for real / near real-time large data analysis in remote sensing applications.

Semantic place prediction problem is the process of giving semantic names, such as school, home, office, to locations. Different than the localization problem where the coordinates of a place are predicted, the aim is to semantically characterize the location. While the GPS coordinates of a place can be utilized in solving the problem, phone usage patterns of the users in that location can be used as well. In this paper, the aim is to semantically classify locations of smart phone users utilizing the data collected from wireless interfaces and the motion sensors available on the phones with machine learning algorithms. The efficiency of features extracted from raw data is analysed in terms of metrics such as accuracy, using different classification algorithms. The results show that, while random forest and decision tree algorithms achieve 66% accuracy with only temporal features, adding features from device properties and activity features increases the accuracy up to 99%.

Artificial Bee Colony algorithm, inspired by the foraging behavior of real honey bees, is one of the most important swarm intelligence based optimization algorithms. Like other population based evolutionary computation techniques, Artificial Bee Colony algorithm is suitable for parallelization on distributed architectures. In this paper, we presented a new emigrant creation strategy that is being distributed between subcolonies running simultaneously on the independent compute nodes. The running times and objective function values obtained by the parallelized Artificial Bee Colony algorithm with the proposed model on different number of compute nodes are compared with the sequential counterpart of the algorithm and it is seen that convergence performance of the parallelized Artificial Bee Colony algorithm is significantly improved with the proposed emigrant creation strategy.

The effect of the unmodeled dynamics and unknown disturbances prevent the accurate control of the real-time systems. The designed controllers must undertake the effect of these undesired uncertainties. In this paper, adaptive uncertainty modeling based model predictive controller is proposed for the control of uncertain linear systems. The uncertainty modeling structure uses an artificial neural network with adaptive learning rate for fast approximation. The stability of the proposed adaptive uncertainty modeling based model predictive control (UMPC) is shown using Lyapunov candidate function. Conventional MPC and proposed UMPC are applied to the control of a real-time DC/DC buck power converter. The conventional MPC cannot accurately control the DC/DC converter due to the unknown parameters and unmodeled dynamics. However, the proposed UMPC controller can accurately control the system with modeling the uncertainties in controller dynamics. The proposed controller is promising to control uncertain systems in future applications.

In this paper, a new flickermeter, which uses lamp illumination as its input and can compute flicker according to the IEC flickermeter standard for any kind of lamp other than the incandescent lamp, is presented. According to the IEC Standard 61000-4-15, flicker is calculated from input voltage and based on the assumption that the input voltage feeds an incandescent lamp. In this study, a system measuring light has been developed and the illumination of three different types of bulbs, with the same input voltage, has been investigated. Based on the results, IEC flickermeter algorithm, which is designed for the incandescent lamp assumption, has been updated to compute flicker any kind of lamp and a new flickermeter system based on the IEC standard has been developed. The developed flickermeter has been tested with the test signals recommended in the IEC standard and it has been shown that the developed flickermeter can accurately compute flicker.

In this paper, a computationally efficient method is proposed for modeling the indoor visible light communications (VLC) channels using a non-sequential ray tracing technique. We created three dimensional realistic simulation environment to depict indoor scenarios specifying the geometry of the indoor environment, the objects inside, the reflection characteristics of the surface materials as well as the characteristics of the transmitter and receiver. We then compute the received optical power and the delay of direct/indirect rays which are used to obtain the channel impulse response (CIR). Finally, LMS (least mean squares) and RLS (recursive least squares) iterative channel equalization techniques which are wildly employed in traditional electrical wireless communications are tested at the receiver of the VLC system. Their convergence rates as well as impacts on the BER performances are investigated by determining the optimal control parameters of these algorithms. It is concluded that for the equalizer operating with its optimal parameters, the BER performances of the VLC systems are very sensitivie to the field of view (FOV) values.

In this paper, Discrete-wavelet transform (DWT) based algorithm are used for speech signal denoising. Both hard and soft thresholding are used for denoising. Analysis are done on noisy speech signal corrupted by AGWN(Additive gaussian white noise) noise. Simulation and results are performed by using SignalLab application which developed for this research. MSE and SNR is used to compare results. Qt/C++ IDE is used to develop the application. This application developed for not only analyzing speech signal, but also can be used in an embedded system or a real-time applications. According to the obtained results, DWT is a good method for speech denoising when parameters are selected correctly.

24th Signal Processing and Communication Application Conference (SIU) (2016 : Zonguldak, TURKEY), Critical Infrastructure Security had always been an issue. With the technological advancements, there has been a continuous inflow of novel methods for critical infrastructure security. In this paper, design, implementation, and field tests of a novel critical infrastructure security system based on wireless sensor networks is presented.

24th Signal Processing and Communication Application Conference (SIU) -- MAY 16-19, 2016 -- Zonguldak, TURKEY, WOS: 000391250900110, In this paper, 11 different power quality disturbances were automatically detected by using statistical features with wavelet transform and norm entropy techniques. The best of the created features were selected with forward selection algorithm. Performance of classification algorithms, Support Vector Machines (SVM), Multi Layer Perceptron (MLP), k Nearest Neighbor (KNN) and random subspace KNN (Sub-KNN) which is an ensemble method, were examined. Consequently, the best classification accuracy of 99.3% was achieved by using Sub-KNN and it was appeared that compared to other methods, this algorithm was more robust against the noise., IEEE, Bulent Ecevit Univ, Dept Elect & Elect Engn, Bulent Ecevit Univ, Dept Biomed Engn, Bulent Ecevit Univ, Dept Comp Engn

In this paper, a new optical orthogonal frequency division multiplexing (OFDM) method is proposed for multiple-input multiple-output (MIMO) visible light communications (VLC) systems. In the new scheme, non-DC biased optical OFDM (NDC-OFDM) and optical spatial modulation (OSM) methods are combined to obtain a novel optical OFDM scheme which has a better error performance. The new scheme is named as non-DC biased optical spatial modulation OFDM (NDC-OSM-OFDM). It is shown via computer simulations that the proposed scheme achieves better bit error rate (BER) performance than DC-biased optical spatial modulation OFDM (OSM-OFDM-DC), asymmetrically clipped optical spatial modulation OFDM (OSM-OFDM-AC) and NDC-OFDM.