ABSTRACT. Advances in networks and computing systems have led to new platforms for computing, such as Clouds. The increasing popularity of cloud computing has offered computational services to many scientists, consumers and enterprises as utilities, on a pay-per-use approach. Many business organizations migrate all or part of their data processing activities in the cloud. In addition to enterprise and scientific computing, clouds are also being used for the management of smart facilities. There are important issues that must be addressed in cloud computing, such as: performance, resource allocation, efficient scheduling, energy conservation, reliability, protection of sensitive data, security and trust, cost, availability, quality. Effective management of cloud resources is crucial to use effectively the power of these systems and achieve high system performance. The cloud computing paradigm can offer various types of services, such as computational resources for complex applications, web services, social networking, urban mobility, health care, environmental science, etc. Furthermore, the simultaneous usage of services from different Clouds can have additional benefits such as lower cost and high availability - however issues like interoperability, portability and service brokerage need to be addressed. Recently, Big Data has become one of the most important research fields in science, engineering, enterprise, biology, healthcare, etc. However, in order that cloud computing will be a platform for supporting big data applications, appropriate algorithms are required for acquiring knowledge from a variety of big and not centrally collected data. In this talk we will present state-of-the-art research covering a variety of concepts on cloud computing, based on existing or simulated cloud systems, that provide insight into problems solving and we will provide future trends and challenges in the cloud computing area.

ABSTRACT. Public transportation billing systems usually relay on different technologies ranging from mechanical systems, through magnetic cards to contact and contactless smart cards. This paper shows a possible solution that could be used to detect a moment when a passenger leaves a vehicle. The solution relays on data gathered from accelerometers in vehicle and passenger’s mobile device, and in that way relieves the passenger from interaction with the billing system. Two approaches for comparison of gathered accelerometer data are considered. The first that compares series of accelerometer vector intensities and the second that compares acceleration gradient. The approaches are evaluated using newly developed system based on Android application.

ABSTRACT. This paper describes the system for collecting data about transport comfort. Those data are received via 9DOF sensor board, which gives Euler angles (yaw, pitch, and roll), as well as the values of 3-axes accelerometer, gyroscope, and magnetometer. These values are intended for use with Android application that collects and displays them, along with the GPS coordinates. Also, the user has the ability to subjectively assess the current comfort level.

ABSTRACT. In order to achieve better performance results for the generated code various compiler optimizations are performed during the compilation process. Even though it is expected that optimization would improve the code quality, for embedded processor with small amount of resources applying certain optimizations can sometimes lead to generation of larger number of instructions (i.e. larger code size and slower execution). One such optimization is a loop optimization - invariant code motion. In this paper we present improvements to the implementation of this optimization in an embedded processor oriented compiler infrastructure which give good results in cases of high register pressure. The enclosed test results for translated code with and without proposed improvements show that our new approach generates code with better performance in all cases.

ABSTRACT. Remote monitoring of beehive colonies becomes an interesting topic in the field of precision agriculture. One of the most relevant parameters which give insight into beehive health is the bee activity at beehive entrance. This paper presents design of a dedicated bee counter which is located at the beehive entrance to count bees entering or leaving the beehive. Paper also presents the counter’s implementation on FPGA platform which can be used for field measurements of the bee activity.

ABSTRACT. Hard disk drives are more prone to failures than other parts of computer systems because they are the only electromechanical component susceptible to mechanical wear. Failure of hard drive leads to permanent data loss, which is typically more costly than drive itself. Prediction of hard drive failure provides notification to user copy stored data to another storage device, preventing data loss. SMART technology enables monitoring of hard disk drive vital parameters and warn used when some parameter exceed manufacturer defined threshold. Advanced failure prediction techniques rely on machine learning to find dependence between multiple SMART indicators in order to provide more precise prediction of hard drive failure. In this paper, we presented anomaly detection model which enabled prediction of hard drive failure on average 39 days prior actual failure of the hard drive.

ABSTRACT. RAID storage technology improves the reliability of the computer storage systems. Hardware RAID 6 systems have been traditionally using Reed-Solomon codes. However, in software RAID 6 systems, different coding schemes can be employed. Parity array algorithms show satisfactory results, as they display significantly higher execution speed. In this paper, we have evaluated parity array EVENODD algorithm as a candidate for RAID 6. We have listed its main advantages compared to the similar algorithms. Additionally, we have shown the performance improvements that can be achieved through its parallelization on general multicore processors.

ABSTRACT. This paper presents a method of realization of the system for accurate measurement of small weights by using a programmable system on chip. The designed system is based on a programmable system on chip CY8C5568AXI-060 of PsoC-5 family and the implementation of a laboratory model of the system was performed by using evaluation board CY8CKIT-050 and load cell with strain gauges CZL601. The implemented system has been tested in a temperature range from 10 ˚C to 50 ˚C and the achieved results show that the accuracy of the system is better than ±0.5% for weights of 0.1 to 5 kg.

ABSTRACT. Sistem za daljinski nadzor i upravljanje uređajima energetske elektronike SDNU prvi put je primenjen pre 12 godina. Do sada je montirano preko 500 perifernih uređaja i formirano 35 centara za nadzor. U toku eksploatacije sistema često je prilagođavan zahtevima korisnika. Od korisnika se pojavio i zahtev da prikupljeni podaci u istom trenutku budu dostupni većem broju korisnika. Pri tome bi grafički prikaz bio prilagođen za prikaz podataka koji su od interesa za svaku grupu korisnika. Zahtev iziskuje reorganizaciju SDNU. Pored izmena u centaru za nadzor neophodno je promeniti i način prikupljanja podataka. U radu je opisano realizovano rešenje.

ABSTRACT. U Laboratoriji za fiziku Instituta za nuklearne nauke “Vinča” u Beogradu, 2016. godine razvijena je i koristi se kontrola za daljinsko upravljanje izvorima za napajanje. Ovi izvori napajaju mnoge elemente za transport jona, skretne magnete, magnete za fokusiranje jonskih snopova i broje preko 40 uređaja a biće opisani na primeru daljinske kontrole izvora francuskog proizvođača Sigmaphi. Izvor napaja jednosmernom strujom solenoidno sočivo SL(1) na kanalu C1. Kanal C1 se koristi za analizu jonskih snopova koji se dobijaju iz mašine M1 koja predstavlja ECR(2) izvor višestruko naelektrisanih teških jona i sa C1 je deo postrojenja FAMA(3) , koje opet predstavlja niskoenergijski deo akceleratorske instalacije Tesla. SL(1) – solenoidno magnetno sočivo, transportni element kanala C1 koji služi za fokusiranje snopa na meti, proizvođač- NIIEFA (Efremov Research Institute, St.Petersburg, Russia) ECR(2) – izvor plazme na bazi elektronske ciklotronske rezonance FAMA(3) – Postrojenje za modifikaciju i analizu materijala jonskim snopovima (“Facility for Modification and Analysis of Materials with ion beams”)

ABSTRACT. U ovom radu je prezentovana efikasna hardverska implementacija dvodimenzionalne direktne diskretne wavelet transformacije (DWT) koja koristi 5/3 filtre kao osnovne gradivne blokove. Predložena implementacija zahteva korišćenje manjeg broja konfiguracionih logičkih blokova u poređenju sa ostalim implementacijama poznatim iz literature. Predložena implementacija zahteva i korišćenje memorije manjeg kapaciteta od bilo koje druge do sada objavljene implementacije. Opisana implementacija ima maksimalnu radnu frekvenciju uporedivu sa drugim implementacijama iz literature, i pritom ima manju potrošnju u poređenju sa njima. Prezentovana implementacija je pogodna za korišćenje u sistemima za kompresiju digitalne slike koji koriste 5/3 filtre, kao što je npr. JPEG 2000.

ABSTRACT. U ovom radu je opisana efikasna hardverska implementacija dvodimenzionalne inverzne diskretne wavelet transformacije (DWT) koja upotrebljava inverzne 5/3 filtre kao osnovne gradivne elemente. Prezentovana implementacija ima manji utrošak logičkih resursa u poređenju sa ostalim implementacijama opisanim u literaturi. Predložena implementacija koristi i memoriju manjeg kapaciteta od bilo koje druge do sada objavljene implementacije. Prezentovana implementacija ima manju potrošnju u poređenju sa drugim implementacijama iz literature, dok je njena maksimalna radna frekvencija uporediva sa maksimalnim radnim frekvencijama drugih implementacija. Predložena hardverska implementacija se može iskoristiti u sistemima za dekompresiju digitalne slike koji koriste inverzne 5/3 filtre, kao što je npr. JPEG 2000.

ABSTRACT. A perfect absorber, i.e., a structure which absorbs 100% of the incident acoustic energy, of very small thickness is of great scientific and engineering interest. Until now, porous or fibrous materials have been the common choice for noise passive control due to their ability to dissipate sound through thermal and viscous losses. This results in limitations: to absorb low frequency sound, bulky and heavy treatments are required even when optimized multilayer or graded materials are used. For many years the development of noise reducing treatments has been the subject purely of acoustics research. However, recent scientific advances provide a unique and timely opportunity to bring about significant improvements in the design of noise treatments. This presentation aims at giving an overview of some solutions that have been recently developed in LAUM to tackle the problem of sound absorption by rigidly backed subwavelength structures.

ABSTRACT. This paper exhibit the research and development process of sound reduction panel for existing walls with the specific market needs. According to the contemporary urban way of life many people in the cities are living in acoustically weak building constructions. Increased noise levels generated from the occupants is provoking annoyance and the need for efficient sound attenuation panels for existing partition elements is growing. Not only the sound reduction properties but the thickness of the panel, simple application and visual properties are from significant meaning. According to those needs was developed a thin layered sound reduction panel with sound attenuation and sound absorption properties.

ABSTRACT. Drum sound effect is common source of noise in the industries. Noise generated by impact of two rigid objects with low internal losses, is characterized by high peak value and short duration. When two bodies collide, their kinetic energy transforms to another form of energy such as heat, plastic deformation, fracture and sound. In order to reduce the drum sound effect, contact surface material is investigated. Materials with different properties and thickens are measured in laboratory conditions. For numerical calculation the Hertz contact theory is used.

ABSTRACT. This paper presents an overview of the research conducted for the purpose of a doctoral dissertation entitled “Multi-criteria Optimization of the Selection of Measurement Strategy for the Assessment of the Long-term Environmental Noise Indicators” by Darko Mihajlov. The dissertation was defended in October 2016 at the Faculty of Occupational Safety in Niš. Based on the results of long-term noise monitoring at chosen measurement points for the purpose of selecting the optimal measurement strategy, the research involved an attempt to maximally reduce the measurement time interval for each measurement point while acknowledging the mutually opposed requirements for sufficiently accurate and precise results of semi-permanent monitoring in relation to the experimental results of permanent monitoring on the one hand, and the degree of utilization of the measuring equipment on the other hand. The optimal measurement strategy for a multi-criteria-defined problem was then selected using the PROMETHEE method, as the most acceptable method in terms of possible choices of how to determine the inter-comparison of alternatives in relation to the nature and values of specific criteria. This involved a selection of a suitable type of preference function and a determination of the values of indifference and preference parameters, which, together with the weight coefficients of the criteria, had a key role in producing the final solution to the problem.

ABSTRACT. Industrial facilities typically include equipment installed (1) in free field at some elevation above the terrain (e.g. wind turbines, vents, exhaust stacks, flares etc.) and (2) on top of floating screens (e.g. standard equipment on steel decking). Noise impacts from such equipment are routinely assessed with the ISO 9613-2 prediction method for the purposes of noise control design and occupational noise studies.

The assessments are made at very short distances thus potentially undermining the stated accuracy of ISO 9613-2 when d<100m. Furthermore, ISO 9613-2 is ambiguous in its definition of source emission properties and ways to account for the apparent increase in sound power of sources located close to reflective surfaces.

A simple test scenario was devised to examine the two available ISO 9613-2 ground effect calculation schemes as implemented in SoundPLAN v7.3. The results indicate that neither of the two methods performs as expected when simulating unobstructed short-range sound propagation over a hard ground and at sharp angles of sound incidence (due to elevated and closely spaced sources and receivers).

The observed inaccuracies may not be critical for large-scale noise mapping scenarios, but can significantly influence the facility design projects where noise impacts are routinely assessed at 1m from the equipment. This is demonstrated with a realistic modelling scenario involving a complex industrial noise source (gas turbine generator) mounted on top of a steel deck of an offshore platform.

ABSTRACT. After decades of research, Fourier transform can be easily replaced with wavelet transform in some applications, especially when frequency resolution problems need to be solved. Wavelet usage nowadays is widespread and it includes signal de-noising, which is the main focus of this research. This paper presents analysis of the wavelet de-noising process implementing different wavelet families and parameters on room impulse responses (RIRs) contaminated with different synthetic white and pink noises. All RIRs used in this process are also filtered in third-octave and octave bands. Special attention is paid to the wavelet parameters that have a major role in the de-noising process. The results show that usage of adequate wavelets and their parameters can provide significant reduction of noise effects, or more specifically improvement of the dynamic range of a RIR and corresponding decay curve.

ABSTRACT. In this paper, FSO transmission of the halftoned image over the Double-Weibull turbulence channel is analyzed. First, halftoning method and algorithm for FSO transmission simulation are explained. Then, transmission of halftoned image is described. At the reception side, we have used an average bit error rate (ABER) for reconstructed image performance measure, as a function of FSO Rytov variance. Obtained simulation results are shown in tables and graphically presented. Finally, performance analysis has been carried out, trough comparing results as a function of Rytov variance (in strong and moderate fluctuation regime)

ABSTRACT. In this paper we propose a procedure for calibration of local oscillator mismatch in a collocated multi-channel receiving system. It is based on a new algorithm for joint relative phase and frequency offset estimation successively applied to pairs of channels, and compensation of oscillator mismatch using these estimated parameters. It is assumed that the receiving channels are time synchronized. The theoretical Cramer-Rao Bound (CRB) for phase and frequency offset estimation is derived. Results of simulations are compared with the CRB. Performance of the proposed calibration procedure is also illustrated by experimental results using software defined radio platforms

ABSTRACT. Since the diversity reception is relatively often used in wireless receivers to improve performance in fading channels, this paper investigates equal gain combining (EGC) diversity receiver performance in fading channel. More precisely, the outage probability of EGC receiver in two wave with diffuse power (TWDP) fading channel will be analyzed. The influence of different fading parameters on the system performance will be considered through theoretical analysis and simulation. Finally, some topics for the future work in this area will be proposed.

ABSTRACT. This paper presents an average bit error rate (BER) analysis of the free-space optical (FSO) system employing subcarrier intensity modulation (SIM) with binary phase-shift keying (BPSK). Intensity fluctuations due to atmospheric turbulence are modeled by Malaga (M) distribution. In addition, pointing errors effect is taken into account. Novel closed-form average BER expression for SIM-BPSK FSO system is derived. Furthermore, dual-branch FSO system is considered, when equal gain combining (EGC) is employed, and average BER expression is derived. Numerical results are presented and validated by Monte Carlo simulations. The effects of atmospheric turbulence and pointing errors parameters on the average BER performance are discussed. It is concluded that employing EGC at the receiver results in the FSO system performance improvement, especially when the channel conditions are favorable.

ABSTRACT. This paper proposes a new structure-criterion switching control method for the self-optimized blind decision feedback equalizer (DFE) which switches operation modes according to the mean square error (MSE) convergence state. The new switching control shortens the blind operation period time of the DFE and, hence, speeds up its effective convergence rate. This is achieved combining the MSE estimate of the DFE’s output and a posteriori error of the DFE’s all-pole recursive filter which acts as the front-end amplitude equalizer during the blind operation mode. The efficiency of the switching control is verified by the software simulator transmitting QAM signals over multipath channels.

ABSTRACT. In this paper the analysis of logatome intelligibility of FM demodulated signal is presented. The subjective tests include 50 listeners and 125 logatomes of CVC (consonant-vocal-consonant) type. The experiments are conducted for two values of deviation frequency: 20 kHz and 5 kHz. The confusion matrices for vowels and consonants in the initial and the final position are given as well. The average recognition rate for logatomes is 93.8% (for deviation 20 kHz) and 93.2% (for deviation 5 kHz).

ABSTRACT. U ovom istraživanju je analizirano izdvajanje obeležja za klasifikaciju modelovanih signala osobe koja hoda sa audio izlaza izviđačkih radara na osnovu modifikovanog spektrograma. Modifikovani spektrogram je računat na osnovu uvijene diskretne Furijeove transformacije, dok je za izdvajanje obeležja ovako dobijenog spektrograma korišćena obrada njegovih binova primenom tehnika digitalne obrade slike. U radu je analizirana mogućnost izdvajanja obeležja u zavisnosti od odnosa signal-šum. Dobijeni rezultati pokazuju da je primenom uvijene diskretne Furijeove transformacije moguće izdvojiti obeležja za klasifikaciju signala sa izlaza izviđačkih radara pri manjim odnosima signal-šum u odnosu na korišćenje diskretne Furijeove transformacije.

ABSTRACT. U cilju poboljšanja karakteristika regulacije temperature peći za zagrevanje autoklava sa uzorcima za izvođenje eksperimenta hidrotermalne sinteze nanomaterijala izvedeno je hardversko poboljšanje i razvijen je softver za daljinsko upravljanje i nadglednje procesa zagrevanja personalnim računarom (PC). Hardversko poboljšanje se odnosi na zamenu postojećeg sistema regulacije termostatom, kojim se vršilo ON-OFF upravljanje sa histerezisom, regulacijom zasnovanoj na PID i ON-OFF upravljanju i uvođenje dvostepenog sistema zagrevanja. Kontroler kojim se vrši upravljanje povezan je sa PC računarom preko RS485 interfejsa. Namenski softver MicroOven je razvijen u cilju zadavanja parametara i praćenja toka eksperimenta u realnom vremenu. Softver komunicira sa udaljenim temperaturnim kontrolerom preko Modbus RTU komunikacionog protokola. Na ovaj način došli smo do uštede od preko 50 % u odnosu na slične komercijalno dostupne proizvode.

ABSTRACT. Detekcija manevra je značajna u sistemima za praćenje ciljeva u vazdušnom prostoru. U ovom radu se cilj prati infracrvenim video senzorom. Snimljene video sekvence su obradjenje i formirana merenja, koja su upotrebljena za izdvajanje obeležja cilja - orijentacija, površina i intenzitet. Predložen je pristup za detekciju različitih manevara cilja na osnovu analize izdvojenih obeležja tokom vremena. Analizirana su dva osnovna manevra cilja, rotacija oko centralne ose i promena stava u ravni. Data je osnova za detekciju složenijih manevara

ABSTRACT. Elektroenergetski sistem, kao izrazito nelinaran i kompleksan sistem, mijenja svoje radne karakteristike kontinualno. Vremenom, kako raste potreba za električnom energijom, uslovljena uvećanjem potrošakog konzuma, rastu i proizvodni kapaciteti. Taj rast dovodi do sve većih zahtjeva u cilju očuvanja stabilnosti sistema. Uređaji za sinhronizovano mjerenje fazora (Synchronized Phasor Measurement Units), pronašli su svoju primjenu u modernim elektroenergetskim sistemima. U ovom radu je predložena metoda za mjerenje fazora i frekvencije, na osnovu mjerenog signala faznog napona/struje. Algoritam se bazira na primjeni Weighted Least Squares Constant Modulus algoritma. Rezultati simulacija pokazuju da predloženi pristup postiže veću tačnost estimacije u odnosu na razmatrane metode.

ABSTRACT. The work considered the reduction of the High Order System Model with or without time delay, in the First Order Plus Time Delay or Second Order Plus Time Delay System Model by applying Skogestad “Half Rule” method. Time and frequency characteristics of the high order system model were compared to the characteristics of approximation models, and the effect of this reduction method on the dynamics of the corresponding discrete models for various periods of discretization were analyzed.

ABSTRACT. U ovom radu biće prikazan postupak formiranja dijagonalne reprezentacije modela jedne široke klase linearnih, stacionarnih sistema koji se opisuju neracionalnim funkcijama prenosa. Prikazani postupak će se u potpunosti osloniti na alate klasične kompleksne analize, koji se tradicionalno široko primenjuju u teoriji upravljanja. U tom smislu, opisani postupak se donekle izdvaja od pristupa koji se obicno sreću u literaturi, a koji se oslanjaju na teoriju uopštenih funkcija (odnosno distribucija). Smatramo da će ovakav pristup učiniti izlaganje pristupačnijim inženjerima. Biće diskutovani uslovi pod kojima je moguće iracionalnu funkciju prenosa realizovati u dijagonalnoj formi; na osnovu toga, biće prikazano uopštenje Hevisajdovog i Pronijevog razvoja.

ABSTRACT. U radu se posmatra vremenski kontinualan multivarijabilni sistem upravljanja sa ciljem njegove stabilizacije u konačnom vremenu. Predloženo rešenje koristi klizni režim višeg reda koji koji anulira red sistema i na taj način se obezbeđuje dolazak trajektorije sistema u koordinatni početak za konačno vreme, uprkos dejstvu poremećaja koji zadovljavaju uslov poklapanja. Predloženi pristup zahteva da je red sistema jednak celobrojnom umnošku broja ulaza. Kada to nije slučaj, sistem se proširuje do potrebnog reda ubacivanjem određenog broja integratora na ulaze sistema. Za tako prošireni sistem data je metoda projektovanja klizne površi koja obezbeđuje željeno ponašanje i predloženo je odgovarajuće upravljanje. Performanse projektovanog sistema su pokazane na simulacionom primeru.

ABSTRACT. Effective utilization of agricultural machinery significantly affects the economy of agricultural production, food security/safety, and environment. In this paper we present a novel approach to creation of field operation lines, which considers imprecisions in data. The operation lines construction bases on fuzzy representation of geo-spatial data, and linear models obtained by bagging and ensemble methods of machine learning. For representing imprecision in geo-spatial data we use a simple but yet efficient data model based on fuzzy linear space. We train our models using the training set consisted of machine positions acquired via GPS device.

ABSTRACT. In this paper one important entirety in most OCR systems is discussed. The main goal is to explain the approach for segmentation of the individual characters from machine-typed documents, which is based on different image processing techniques in spatial domain and exploits the features of a document structure. The paper gives the overview of the proposed technique by the phases, starting from the grayscale conversion and binarization, through the line, word, and character segmentation of a document image using the modified projection profiles technique, as well as the description of the problems which can appear. The approach is semi-automatic since it requires adjustment of the thresholds for the different segmentation parameters. The numerical results which show the proposed approach performances from the perspective of the time complexity and segmentation accuracy are provided. Results show that this approach outperforms the state-of-the-art methods in all aspects. Furthermore, it is shown that the proposed character segmentation technique is suitable for real-time tasks.

ABSTRACT. In this paper, we investigated a way of computing data, encrypted using Paillier’s homomorphic encryption scheme and stored in Cassandra database. For computing data, we used Cassandra’s User-Defined-Functions (UDF) and User-Defined- Aggregates (UDA). Sensitive data, such as customers’ consumed energy was used as an example, where aggregated energy consumption, over a period time, was computed using UDF and UDA. Cassandra database was deployed on Microsoft Azure platform, where six different clusters were set, consisting of two, four and six nodes.

ABSTRACT. The problem that we tried to solved is mathematical modeling of the system for precise prediction of appropriate time for successful chemical protection of fruits. Precise prediction based on the use of mathematical modeling imported in software applications will reduces number of chemical treatments, provides economical savings for farmers, cheaper food for customers, and finally, and most importantly to reduce pesticide residues in fruits. We use mathematical modeling in order to create patterns between weather parameters, parameters about pathogenic spore activity and to find dependencies between occurrences of diseases and observed parameters. Created mathematical model provides good dataset visualization, outlier detection, and accurate prediction in 89% of applied tests.

ABSTRACT. The research presented in this paper refers to an approach to solve the Knapsack problem using Upgraded Petri Net (UPN). The problem may be pointed to as the selection of the cars featured with length and price resulting in settlement of appropriate cars in the truck container of given length. Models refer to appropriate genetic algorithm extended with the correctness function. This paper describes the simulation and analysis of the UPN models. One should state that previously developed suite of Petri Net Manager (PeM) is used to simulate and analyze the models presented in this paper.

ABSTRACT. In this paper a novel - simplified method for calculating energy efficiency, and it’s implementation were presented. Technical details of the method were explained. It was shown how this software should be used, and how results should be interpreted. In this way, the use of software for energy efficiency calculations was simplified, and that was achieved with minimal loss of accuracy and precision. The results were compared with the results from a similar software tool – used in Serbia, using concrete data from location studies. Possibilities for further improvement were suggested.

ABSTRACT. The different approaches to the processing of negation in sentences written in Serbian and their influence on the sentiment analysis are presented in this paper. Sentiment analysis is still not sufficiently elaborated, especially for resource-limited languages. Processing of negation is important for sentiment analysis and significantly increases its quality. The existing solutions for the Serbian language which describe the sentiment analysis have not quite provided an appropriate and efficient way of processing the negation in sentiment analysis.

ABSTRACT. The paper presents the academic framework for curricula development in the Erasmus+ project on Master Studies in Digital Broadcasting and Broadband Technologies (DBBT) at Electronics and Electrical schools of six Western Balkans universities and with the participation of three European Union universities.

ABSTRACT. Highly linear and low-noise transimpedance amplifier (TIA) based on CMOS inverter as an amplifier within multiple feedback (MPFB) filter topology is presented in this paper. Absence of complex analog circuitry and compatibility with widespread digital CMOS processes highlight this TIA topology as the best choice in signal conditioning chains approaching ADCs. Special care is devoted to TIA stability and compensation networks. Variations in TIA performance induced by supply noise is made more robust using dedicated low-dropout (LDO) regulator. Transimpedance and bandwidth programmability is ensured by design of the feedback network. Nominal transimpedance and bandwidth at 65 0C are 72 dBΩ and 5 MHz, respectively. Achieved third-order intermodulation distortion is -84 dBc, while input third-order intercept point is 34 dBm. This performance is comparable with state of the art TIA solutions used in optical receivers and communication baseband circuits.

ABSTRACT. In this paper we present analysis of area efficiency of conventional binary-weighted and split-capacitor topologies of switched-capacitor DAC for SAR ADC. Although the main reason for usage of split-capacitor topologies is reducing the DAC’s area, the analysis showed that it is not always the case since the linearity parameters of split-capacitor topologies are more sensitive to parasitic effects. We based our analysis on differential realization of 12-bit switched capacitor digital-to-analog converter for 1 MS/s differential SAR ADC in 180 nm CMOS. However, the analysis can be applied to any SC DAC used in SAR ADCs. We used common mode voltage as a third reference voltage potential besides GND and VREF, which allowed us to reduce the DAC’s area while maintaining the linearity specifications. Monte Carlo simulations of the designed DAC with extracted parameters gave the 12-bit performance at 1 MS/s with the differential nonlinearity in LSBs DNLmax,mean/σDNLmax = +0.395/0.201 and integral nonlinearity INLmax,mean/σINLmax = +0.326/0.108.

ABSTRACT. This paper evaluates the design method of short wordlength FIR (finite impulse response) filters. The design is based on the change of basis of polynomial space, which means that a filter is designed as a weighted parallel structure consisting of FIR filters with only 1s and 1s as coefficients of the transfer funtion, i.e. only adders/subtractors and unit delay lines in hardware implementation. The structure allowed for a higher minimal attenuation in the stopband in 87.8% of designed filters. Negligible errors in the cutoff frequency and passband attenuation were noticed. Also, this paper broadens the existing design method to high pass, bandpass and bandstop filters. Finally, IIR (infinite impulse response) filters were discussed as a possible generalization of the design procedure.

ABSTRACT. In phase (I) and quadrature (Q) complex signal representation (I+jQ) of the modulated signal is widely used in RF systems. Ideally, I and Q signals have equal amplitude while Q is 90o phase shifted relative to I. Having quadrature signaling provides many advantages such as higher RF spectrum efficiency (more bits/Hz), lower data converters (ADCs/DACs) sampling rate for the same data throughput, computational power of the base band modem is relaxed, all making low power transceiver implementations possible. However, any mismatch of gain or phase between I and Q (IQ imbalance) will degrade transceiver performance. Error vector magnitude (EVM), bit error rate (BER) and RX sensitivity in homodyne radios and TX spectrum re-growth in heterodyne architectures are among the most critical parameters affected. There are multiple sources of IQ imbalance. RF mixers can have different gain for I and Q paths, and the gain can also be frequency dependent. PLL which is responsible to generate quadrature LO produces nonequal I and Q signals in terms of amplitude as well as phase error. Even base band modules such as low pass filters or gain stages can contribute. The most difficult to estimate during the system design phase is in fact IQ cross talk happening both on the PCB and inside the RF IC. This paper provides insight into an adaptive filtering method for the correction of the I/Q imbalance in the receive chain. The image rejection is evaluated using 64- QAM and 16-QAM test signals for different adaptive filter designs. The efficiency of the algorithm is inspected in the parameter space consisting of the adaptation step-size and the order of the adaptive filter. Simulation results show that zero order filter and reasonably low value of step-size provide optimum filtering for the above mentioned test cases. Constellation diagrams of the simulated received and adaptively filtered signals illustrate the importance of the adaptive filter key parameters setting.

ABSTRACT. This paper presents design procedure for efficient design of IIR digital filters with critical monotonic amplitude characteristic (CMAC). Design methodology is based on traditional approach i.e. synthesizing the filter transfer function in s-domain and utilizing appropriate conformal mapping to produce z-domain transfer function. Only hardware, fixed-point, implementation is considered. Presented procedure covers design flow from system level to the register transfer level (RTL) modeling and simulation. As an example, highly selective, notch, filter is designed. Obtained simulation results confirmed proper functionality of the designed filter.

ABSTRACT. Inter-chip communications often exploit the so-called serial peripheral interface (SPI) bus for lower-level and medium-speed exchange of data. Typical use cases include application-specific integrated circuit (ASIC) setup and testing via scan chain. A lack of formal standardization and ubiquitous presence inevitably lead to abundance of SPI protocol definitions. This paper presents a custom SPI bus slave controller which includes single and block addressable bank of read-only status, as well as read/write configuration registers. The controller employs a rather efficient in-frame addressing scheme which is supported by low-propagation register bank design. Thanks to these two, the FPGA bus implementation managed to be clocked at 100 MHz, thus effectively achieving a 100 Mbit/s burst data throughput.

ABSTRACT. Acoustic camera represents a modern tool in acoustic measurements. Many measuring microphones in this measurement system would significantly increase the price of the system. Measuring microphones nearly have a flat amplitude characteristic and a linear phase characteristic. The motivation for this research was to cut down the cost of the acoustic camera system realization, which was expensive do to the large number of microphones used. The idea was to use microphones of lesser quality compared to the measuring microphones and with specific digital processing get a resulting signal that is comparable to the signal we got with the measuring microphone. In this paper, an equalization technic was used with the help of an adaptive algorithm to improve the frequency characteristics of the microphone of lesser quality. Results of the experiment show that this method can lead to budget cuts and still preserve the accurate location of the sound sources.

ABSTRACT. A design of an electroacoustic instrument prototype for localization of a dominant noise source in the living and working environment is presented in this paper. Presented prototype is based on the conventional delay-and-sum beamforming algorithm which is implemented on a Field Programmable Gate Array platform, (FPGA). Sound signals are provided by receiver system, based on a digital MEMS microphone array. The circular array is oriented in a horizontal plane, enabling 360 degrees of acoustic viewing angle. Configuration enables the implementation of the beamforming algorithm in real time. Additionally, FPGAs brings many benefits in terms of safety, reliability, rapidity, and power consumption. The system has been designed using rapid prototyping methodology with Matlab Simulink tools. Prototype was build, tested and used to validate numerical simulations. A performance evaluation of the FPGA design is presented in terms of hardware resources for the chosen Xilinx Spartan-6 family. Realized prototype was used for field measurements of heating plants in Novi Sad and Belgrade. Results of measurements were successfully applied in definition of sound sources and model, providing the best background for optimal noise mitigation measures.

ABSTRACT. This article describes the implementation of a MATLAB developed calculation algorithm used for sound scattering investigation. The investigation has been conducted by using standard staggered grid leapfrog Finite Difference in the Time Domain (FDTD) numerical scheme in two dimensions. The shape under investigation is single plane Schroeder sound diffuser. The goal of this article is to apply the Finite Differences in the Time Domain numerical method for objective measurement of sound scattering in enclosed spaces and thus aiding the design and optimisation of room acoustics in such spaces.

ABSTRACT. Noise control in open offices must be focused at reducing disturbance noise which will increase the speech privacy. There are no regulations about the acoustical design of open offices in Bulgaria. Usually the room acoustics is controlled by techniques high room absorption, desk baffles, office furniture, workstation positions and masking sound. The interaction between these techniques is complicated, because the speech privacy is studied at different distances from a speaking human. The aim of this study is to evaluate the speech transmission index and A-weighted speech level when sound masking system is available and applicable in open space premises. The examined model is created in real open space environment and the measurement results are obtained from sound analyzer which will give the degree of the effectiveness of sound masking systems. Recommendations for the acoustic ambience of open offices are made suggested.

ABSTRACT. The microphone, amplifier and photoacoustic cell affect the transfer function of the gas-microphone detection of the photoacoustic measuring system. Experimental determination of the transfer function of the system detection part eliminates the influences of these components occurring due to the system design. Based on the measurements performed in an anechoic chamber where the swept sine signal was used as an excitation, the transfer function is determined. Two different acquisition cards were used for the signal detection. One of them was used to generate the excitation signal simultaneously with the microphone signal detection. The approximately flat transfer function up to 6 kHz is obtained with that card, which is not the case with the other acquisition card. In the latter case, the determined transfer function shows the oscillations in the regions where the signal to noise ratio is insufficient.

ABSTRACT. Traditional measurements of the transmission loss (TL) has been performed for many years using the proper infrastructure. This requires accessibility to the specialized facilities such as reverberant and anechoic chamber, which is available only to a small number of laboratories throughout the world. In order to broad the possibility of transmission loss measurements to a wider population of acousticians, scientists are searching for new measurements methods and algorithms. This paper exploits a method for TL calculation using modification of a standard impedance tube, usually used for absorption coefficient measurements. Focus has been paid on the design of the tube, as well as evaluation through repeatability of measurements. In addition, testing has been performed on the one sound insulation sample and results of the TL calculation discussed.

ABSTRACT. This paper describe the Android application for recognition of arbitrary sounds. Application uses algorithm previously used for automatic speaker recognition. First 18 mel-frequency cepstral coefficients are used as sound features. Twenty exponential auditory critical bands are used during their determination. Covariance matrices are applied for modeling of sounds. User of this application creates own sound database. Application supports capture of sounds in mono wav format, 16 bit resolution, with frequency sampling of 22050 Hz. After sound capturing is finished user can choose between training and testing procedure. Training is devoted to creating of new model for observed sound and saving in sound database on memory card of mobile phone for example. Testing procedure run modeling of observed sound and compare that model with reference models in sound database. As result of recognition application writes the identity of most similar reference model and also the second and third most similar model and measures of difference with respect to them.

ABSTRACT. This paper presents the results of normal and whispered speech recognition using RASTA (Relative Spectral) filtering applied on PLP (Perceptual Linear Prediction) feature. It is used for a front-end while the DTW (Dynamic Time Warping) algorithm is used for a back-end of ASR system. All isolated words in these experiments are from the Whi-Spe database. In the experiments four training/test scenarios are analyzed: normal/normal, whispered/whispered, normal/whispered and whispered/normal. The speaker dependent mode is used with 10 speakers. The results confirm a good improvement in recognition when RASTA filtering is applied, especially in the mismatch scenarios.

ABSTRACT. Software switches can provide high level of flexibility and are commonly used in software defined networks. In recent years, development of software routers has shifted towards fast network I/O platforms such as netmap and DPDK that bypass slow kernel network stack. Designing and developing software switches is challenged by network requirements which in many cases may be conflicting. Recently, VALE software switch based on netmap platform has been improved with extension called mSwitch. The VALE architecture is redesigned and divided into two fundamental parts: switching fabric that forwards packets and switching logic module that determines destination port. In this paper, we present the mSwitch (VALE) architecture and guidelines for developing switching logic module. Through extensive tests, we determine performance of packet forwarding by mSwitch.

ABSTRACT. Modern packet networks must support high speed packet transmissions. Routers and switches have a major impact on the packet network performance. Several high performance packet switch architectures have been proposed in the literature. One of the most popular packet switch solutions are load-balanced Birkhoff von Neumann switches (LB-BvN) which comprise two switching stages. LB-BvN switches are very popular because the configurations of the switching stages are deterministic, hence it is not necessary to recalculate switch configurations in real time. This property enables high scalability of the LB-BvN switch architecture. Very important feature of packet switches is fair service when some of the output ports are overloaded. The switch should provide fair share of overloaded port's capacity to all flows destined to the overloaded output port. This paper analyzes the fair service of the most popular LB-BvN based switches.

ABSTRACT. The constant development of communication networks imposes big challenges when providing necessary performance and keeping costs low. Successful network and device design rely on testing tools which are capable of simulating real network conditions. Normally, these tools utilize various commonly used IP traffic models in order to generate IP traffic. However, majority of those IP traffic models are useful for specific traffic scenarios and fail to adequately simulate wide range of different traffic patterns. In this paper, we introduce new approach for IP traffic generation based on hidden Markov models (HMM). The important feature of hidden Markov models is their ability to automatically adjust parameters using Baum-Welch algorithm, which allows us to record real IP traffic and dynamically create accurate representation of that traffic. The principles of hidden Markov models and Baum-Welch algorithm are presented and analyzed in details. We develop application which utilizes hidden Markov models that is capable of generating a large number of different IP traffic patterns.

ABSTRACT. This paper presents simulation programs for four traffic models, implemented in mobile telephony. These models differ according to the traffic type in the mobile cell. The first one is the ordinary model, where only external connections exist. The second one is model, which includes some percent of intra-cell connections (internal connections between two users from the same cell). In the third model two types of connections are established: full-rate connections, which seize one traffic channel and half-rate connections, which seize half a traffic channel. And, finally, the fourth one is model with intra-cell and external connections, but with limited number of mobile users in the system. Flow-charts of these simulation programs are explained. The simulation programs are originally developed, based on simulation programs for switching systems, which are also developed in IRITEL. The main purpose of simulation programs is to approve the results, obtained by mathematical analysis. In complicate situations, where mathematical equations for calculating important system parameters may not be easily determined, simulation results may replace calculation.

ABSTRACT. This paper examines the potential of low-cost commodity computers for hosting network of middleboxes and switches, and providing flexible inexpensive platform for packet processing and filtering. The performance of one such solution is reviewed.

ABSTRACT. This paper focuses on the determining the position of a target by measuring the time difference of arrival (TDOA) from a set of receivers whose positions are known. In the first step, weighted least squares (WLS) based optimization function is proposed to locate the target. In the second step, the TDOA target localization problem based on least squares (LS) is formulated as an optimization problem, with a highly nonlinear and multimodal objective function. There are several algorithms to estimate the position of a target. In this paper, we proposed a hybrid Genetic algorithm (GA) that combines the global search and local search abilities in an effective way in order to improve the performance and the solution accuracy. Gradient based local search algorithms such as the Newton-Raphson and Gauss-Newton are the most common methods used for hybridization. The paper investigates the performance of the hybrid genetic algorithm in modelling a TDOA positioning problem using a different local search optimization methods. The simulation results illustrate the performance comparison of these different optimization algorithms.

ABSTRACT. This paper presents and evaluates a Sequential Monte Carlo approach for target estimating. In a surveillance situation, the origin of each measurement is uncertain. Each measurement may be a false (clutter) measurement, or it may be a target detection. Probabilistic Hypothesis Density methods are usually used to discriminate between the clutter and the target measurements. Clutter measurement density is an important parameter in this process. The values of the clutter measurement density in the surveillance space are rarely known a-priory, and are usually estimated using sensor data and track information. This method enhances target tracking performance, compared with standard particle filter. Simulation results validate this approach.

ABSTRACT. Object tracking based on video processing has being an active research field for decades. This interest is motivated by numerous applications. Our focus of attention is on traffic control applications where the Kalman filter has shown as a powerful apparatus in tracking of vehicles on the road. This paper is basically oriented to the specific situations where the proper estimation of vehicle motion has to be used due to temporarily occlusion of some objects, in order to keep the information about the motion of occluded object. The proper criteria related to the switching between measurements and predictions are specified and tested in some of typical road scenarios.

ABSTRACT. Interest for speaker recognition increased again with appearance of new opportunities for applications. The goal of this paper is to examine reliable speaker recognition system and to examine influence of different parameter values. The system consists of Gaussian Mixture Model (GMM) based classifier which uses Mel Frequency Cepstral Coefficients (MFCC) as features. Tests were conducted on Temporal Voice Idiosyncrasy (TEVOID) database. First we performed test using parameters recommended in literature and established the baseline system with excellent 99.5% of recognition rate. We examined the varieties in following steps of process: in speech preprocessing, MFCC calculation, GMM training and in training/testing setup. All the tests were performed using only 10 sentences in total for training and testing. The results of these tests are close to saturation, which makes them not robust enough for general conclusions, but useful for further development.

ABSTRACT. This paper proposes a new distributed asynchronous algorithm for drift correction in time synchronization for networks with random communication delays, measurement noise and communication dropouts. Three variants of the algorithm are proposed, based on different current local time increments. Under nonrestrictive conditions concerning network properties, it is proved that all the algorithms provide convergence in the mean square sense and with probability one of the corrected drifts of all the nodes to the same value (consensus). Asymptotic rate of convergence of the algorithms is formulated. It is shown that it is possible to achieve convergence to the common virtual clock. Simulation results give an illustration of the properties of the algorithms.

ABSTRACT. Correct association of observations with objects is one of the most important and difficult tasks in algorithms for tracking multiple targets in the presence of false observations. We consider one modification of the standard Kalman filter which aims to reduce the tracking error by explicitly taking into account the fact that the probability of correct association is less than one. Through computer simulations, we analyze the performance of this method and assess if, and under what conditions, it can improve upon the standard Kalman filter.

ABSTRACT. In this paper problem of distributed multi-target tracking in large scale camera networks is discussed and a new algorithm based on consensus is proposed. The considered networks are characterized by sparse communication and coverage topologies, as well as the presence of clutter and multiple targets, which make the application of distributed frameworks challenging. The proposed algorithm uses Joint Probabilistic Data Associations (JPDA) methodology to address the clutter and multi-target problem. It applies a novel multi-step consensus strategy which adapts network communication weights to reflect differences between the nodes caused by the adopted network setting. Compared to the existing state of the art algorithms, in spite of having reduced communication requirements, the proposed scheme achieves the same performance quality in terms of the mean estimation error and much better performance in terms of the disagreement between the nodes' estimates due to its specific consensus design.

ABSTRACT. The paper describes the regulation by using proportional-integral controller on the first order systems with time delay, where the focus is given to check the stability of the resulting system. Design controllers based on the application of Lambert W function and setting the dominant poles in the desired place in the infinite spectrum of poles system with time delay. The result was obtained by simulation and compared with other well known methods for the design of this controller.