نشریه صنایع الکترونیک
سال یازدهم شماره 2 (پیاپی 42، تابستان 1399)

Recently, a number of Extreme Learning Machine (ELM) based training algorithms have been introduced for training deep neural network structures. ELM based Auto-Encoder (ELM-AE) is one such algorithm that has been used for making multilayer structures and tuning parameters of each layer. In a simple ELM-AE training algorithm, the weights of the first layer are initialized randomly. This issue is a leading factor in producing reconstruction error. The frequent use of ELM-AE in deep network layers results in propagating such errors through deep structures and in decreasing performance as a consequent. In this paper, we introduce a multilayer structure and a new learning algorithm to train it that prevents error propagation. In order to boost the performance of the model, the parameters in the first layer are initialized by a novel type of ELM-AE called Repeated-AE (RAE) rather than by a random selection method. This RAE-based technique determines the parameters in the first layer far better than do the other ELM-AE existed methods. Next, a single hidden layer ELM is applied for handling the classification task. Experimental results for data classification show that the proposed method outperforms some other methods in terms of the average accuracy over all datasets by amounts of 4%, 26%, 17% and 31%. Eventually, so as to verify the performance of the proposed multilayer ELM-AE in application, we used this model to reconstruct images. The reconstructed images obtained by our approach appeared visually a lot better compared to those obtained by the other methods do.

Transmission opportunity and contention window are two important parameters in media access in IEEE 802.11e-based networks. and determine how to access the channel, as well as the length of time the channel is used. In these networks, each wireless station determines its media access probability by adjusting its contention window size. After winning in the contention, with respect to the duration of its transmission opportunity, transmits its data packets without re-contention. In multi hop networks, because of long distances between the source nodes and the destination nodes, the middle nodes are used to relay the data. Therefore, the middle nodes, in their transmission and receiving range, see more nodes as neighboring nodes, which makes it possible to increase traffic intensity and higher collision rates than other nodes. In this paper, the length of the queue, data arrival rate and the amount media access delay are exploited in order to a dynamic setting of the transmission opportunity and contention window to reduce the rate of packet loss and end to end delay.Simulation results show the proposed method improves media access and reduce the amount of packet loss and delay.

Using the ninth pulse of enhanced Loran (eLoran) chain, it is possible to send various messages through Pulse Position Modulation (PPM) and Reed-Solomon (RS) coding. The eLoran receiver should be able to estimate the location of the ninth pulse with an error of less than 1 microsecond, to extract the message. In order to estimate the location of the ninth pulse in the receiver, the method of third zero crossing detection of eighth and ninth pulse chains, like Half-Cycle Peak Ratios (HCPR), are commonly used in the receiver. In this paper, an enhanced algorithm has been proposed to improve performance of the Loran data channel in the AWGN Noisy environments. In this algorithm, instead of using two pulses, each nine pulses in the GRI chain are used to estimate the location of the ninth pulse. Simulation results show that the efficiency of proposed algorithm is better than existing methods of HCPR, especially at low SNR channels.

Accurate mobility models are crucial to suitably model different types of FANETs’ movements, including both rapid and sporadic changes in directions and movement in straight lines. Due to its capability to model the mobility of flying ad-hoc nodes in three dimensions, Gaussian-Markov mobility model has been examined in this work. Our analysis covers two parameters with the greatest impact on the performance of the network; the regularization factor which determines the impact of memory when updating the variables and the time step, which expresses the frequency of updating the parameters. Two commonly known protocols, OLSR and AODV from proactive and reactive types, respectively, have been used to examine the network performance in terms of mean transmission delay and packet delivery ratio (PDR) while the Gaussian-Markov is being applied on the nodes. It is shown that as the regularization factor increases and nodes move more randomly, the performance of both types of protocols degrades in terms of both delay and PDR. For small values of alpha (alpha < 0.6 ), the delay remains almost unchanged in both protocols when the variables are updated frequently, e.g. ts = 0.1 s. The PDR is also shown to decrease as the nodes travel in faster speeds (30 m/s) for all values of . The reactive protocol delivers better network performance when more flying nodes are deployed as relay nodes; however, the proactive protocol fails to achieve the same results due to the redundant control messages passed through the network.

Coriolis Vibratory gyroscopes are one of the most modern types of gyros that has taken the common gyroscopes residence with some difference in the test mass design and elastic suspension test. According to the important features observed in the capacitive excitation of the actuators regarding the piezo-electric actuators, the operation principles and their formulation are completely changed, which require both two dimensional and finite element analysis to evaluate their optimal performance. Due to the sensors are usually vibrating continuously while operation, in this paper a general formulation is presented so that it can fully describe the influence of the parameters related to different frequency operating modes. The main idea of the vibration gyroscope is to replace the rotational rotor with a vibrational structure and to use the effect of Coriolis force, which causes the secondary motion of a sensitive mass to match an angular velocity. In this paper, the analysis of the sensitivity of hemispherical vibrational gyroscopes is discussed. The electrodes location variation, the sensed voltage around the resonance frequency and pick-off/forcer materials will also be investigated. Finally, the results of simulated resonance frequencies are compared and validated with the theoretical principles.

This paper، present a new isolated multi-output converter with simultaneous buck، boost and inverted outputs. Using the single inductor multi-outputs applications and combining it with flyback structures، the proposed converter can produce different isolated output voltages. In present study، the mathematical relations between output voltages and input have been calculated. The proposed converter has been modeled based on signal flow graph approach. Derivations of large-signal, small-signal and steady state models are demonstrated by considering a single inductor four outputs buck/boost and invertd convert. In this paper, a multi output converter in continuous conduction mode (CCM) are modelled based on SFG method. A prototype has been made which generates two isolated boost voltage، ±15V، two isolated buck voltage، ±5V، from single 12V input supply. Simulation and experimental results are presented to proof the concept of the proposed converter and its model in steady-sated، small signal and large-signal behavior. The simplicity and systematic procedure presented in this work show that the SFG can be a good alternative for modeling switching converter with .complicated structures

The motion control systems based on hydro motors, due to the nonlinear and time variant behavior, are complex systems for accurate control. IN this paper, an adaptive control scheme based on self-recurrent wavelet neural network is presented to control dual drive electro servo-hydraulic system. The proposed control scheme does not require the exact dynamical model of the system. As a result, this method will be appropriate for severe nonlinear systems such as electro servo-hydraulic system. In addition to identifying the dynamic of the system, the proposed neural network is trained to predict one-step ahead of the system output. The training process is implemented by employing a novel optimization algorithm called flexible particle swarm optimization (FPSO) algorithm. In order to enhance the performance of this system, two electro servo-hydraulic systems are connected to each other to increase the bandwidth. The controller provided for the electro-servo-hydraulic system is simulated regarding to the unknown dynamic of the system, and the simulation results confirm the effectiveness of the proposed control scheme.

In this paper, one of the applications of MEMS technology in the field of telecommunication systems is introduced. For the first time, by using a variable MEMS capacitor in the CSRR resonator, a new method for adjusting and multi-banding telecommunication resonators is provided independently and simultaneously. In addition, the CSRR resonator is implemented on a HMSIW line and fabricated with the integrated circuit fabrication technology in order to reduce the noise caused by incomplete connection as well as area and loss and to integrate with the MEMS structure. Also, in this paper, the frequency behavior of structures was investigated by magnetic simulation with HFSS software. The results indicate that the transmission loss of the proposed structures is less than 1 dB and the return loss is greater than 15 dB. The proposed resonator works in the S-band frequency with the same frequency behavior as PCB board while its area is reduced by five times compared to the conventional structure by choosing the appropriate substrate for the integration. The proposed fabrication process for the adjustable and dual-band structure is presented which is based on existing integrated circuit technology.

Cryptographic algorithms are undeniable components in providing communication security. On the other hand, to ensure the security of an algorithm, its security against various attacks must be evaluated. In addition to evaluating the security of cryptographic primetives against statistical attacks such as linear cryptanaylsis, differential cryptanalysis and etc., security of cryptosystems against side-channel attack is always a concern. Among side-channel attacks, fault injection-based attacks received extensive attention. Among fault injection attacks, differential fault analysis is considered in this paper. Although differential fault analysis has been applied to many block ciphers, the security of a class of block ciphers known as ARX ciphers has received less attention. In this direction, in this paper, we analyze the security of an ARX block cipher known as MARX-2 against differential fault attack. As a result of this attack, we recover the n-bit key of the cipher using n-bit fault injection. The applied attack is the first differential fault analysis of MARX-2 block cipher, to the best of our knowledge.

Hyper Spectral Images (HSI) collect a lot of information in hundreds of narrow spectral bands. This type of image has been more useful for a wide range of applications in ground surface identification. Here, there are some processes to achieve that proper information. So, finding a way to gain the best accuracy for collecting data has become an interesting field for scientists. As a result, in this paper, we introduced object-based Feature Extraction algorithms (FE) to find out such useful information. The proposed algorithm has four fundamental phases. In the first stage, we use an unsupervised FE such as the PCA algorithm to extract the most significant features of the image. Then, the Gabor filter would add to obtain the local features. In the third step, we use the K-means algorithm to make a segmentation map of the image. Finally, in the last stage, by considering the coordination between pixels of each region and the effects of local relations among neighbor pixels relating to the same object in the image by an appropriate transformation, a function introduced. As a consequence of all these stages, some important and efficient features of the proposed data would extract.

Breast cancer is one of the most common malignant tumors and the main cause of cancer death ‎among women worldwide. The diagnosis of this type of cancer is a challenging problem in cancer ‎diagnosis researches. Several research before have proved that ensemble based machine learning ‎classifiers are able to detect breast cancer spot more accurate. However, the success of an ensemble ‎classifier highly depends on the choice of method to combine the outputs of the classifiers into a ‎single one. This paper proposes a novel ensemble method that uses modified differential evolution ‎‎(DE) algorithm generated weights to create ensemble of classifiers for improving the accuracy of ‎breast cancer diagnosis. This paper proposes an ensemble-based classifier to improve the accuracy ‎of breast cancer diagnosis. As the performance of DE algorithm is strongly influenced by selection ‎of its control parameters, local unimodal sampling (LUS) technique is used to find these parameters. ‎The two most popular classifiers support vector machine (SVM) and K-nearest neighbor (KNN) ‎classifiers are used in the ensemble. The classification is then carried out using the majority vote of ‎the ensemble. The accuracy of the presented model is compared to other approaches from literature ‎using standard dataset. The experimental results based on breast cancer dataset show that the ‎proposed model outperforms other classifiers in breast cancer abnormalities classification with ‎‎99.46% accuracy.‎

In satellite transponder design, both the group delay of demultiplexer and multiplexer filters and the AM/PM distortion due to the high power amplifier (HPA) nonlinearity are critical specifications, which should be in an appropriate range to ensure the required BER for having acceptable satellite service quality. Practical problems that inevitably arise in the manufacturing process of a transponder lead to a slight deviation in the fulfillment of the maximum allowable group delay and AM/PM distortion. In this paper, the acceptable range of the group delay and AM/PM distortion is investigated simultaneously so that the required BER of DVB-S2 service is guaranteed. Since there are no proper analytical relations to achieve this goal, these nonlinear effects are simulated in MATLAB SIMULINK for both the QPSK and 8PSK modulation schemes. It is inferred from the achieved results that the acceptable phase shift and the group delay of the 8PSK modulation is less than those of the QPSK modulation.