نشریه مهندسی برق
سال چهل و سوم شماره 1 (پیاپی 65، بهار و تابستان 1392)

Transmission Control Protocol (TCP) and other its well-kNown implementations without attend to reason of packet loss, consider it as a sign of network congestion and run congestion control commands. This issue is source of performance degradation for TCP in Wireless Ad Hoc networks (WANETs). If we could distinguish between congestion loss and link error loss or use another strategy for changing congestion window, performance would be improved. In this paper, we are presenting a new mechanism to enhance the performance of TCP in WANETs namely TCP-WANET. In particular, we employ an estimation-based method for setting of congestion window size. TCP-WANET is based on InterACK, which is the time between two consecutive ACKs. Extensive packet level simulations, implemented by using ns-2, show that usage of such congestion control protocol can lead to improved performance comparing to the standard TCP protocol.

In this paper, a method for islanding detection based on analysis of transient state signals is provided. K-Nearest Neighbors (KNN) is trained for classifying the transient events. The required features for classifying are extracted through DWT of signals. The proposed method is then simulated on a medium voltage distribution system of CIGRE with two kinds of DGs using DIgSILENT, MATLAB and WEKA softwares. By analysis performed on type of input signal, type of mother wavelet and required transform level, among 162 relay designs, an optimum relay was selected for DGs based on accuracy, speed, simplicity and cost parameters. By evaluation, it was determined that using only one input (voltage) signal not only improves speed and simplicity and reduces costs, also makes accuracy of the proposed relay better than other intelligent and passive methods.

In this paper, the dynamic Bayesian network for modeling behavior has been used for door-crossing behavior. This network easily trains and it supports dependencies between of data’s sensors and the desired behavior of robot on ideal environment and it detects error on data’ sensor by using data based method in real environment. Elimination of undesired sensor performs with structure changing base on interpretation of sensor’s data. Evaluation is based on robots with artificial error in the readings of the sensors. Proposed model for the "door-crossing", which is based on dynamic Bayesian networks, has been able to resist much higher than the model based on Bayesian networks (fixed structure), in the face of adverse and multiple interpretations of the constant error, drift, also cross-talk, random movement and slight changes in the environment map. This model is much resistant by adding maneuver part to it on dynamic environments. The performance of the proposed model on the simulation environment has been evaluated.

Using the gait as a biometric for authentication is a study of high importance in the field of machine vision in recent years. The proposed methods in this area are divided into two categories of model-related and appearance-based methods. In this paper, both of the mentioned methods of feature extraction are used for authentication purposes. In this paper, silhouettes are used for recognition. These images possess a high dimension which brings about complexity of computations. Therefore, the intrinsic dimension of images which is the minimum number of required parameters for image identification is calculated. Then, the LLE algorithm is utilized for dimension reduction of images to a manifold with an intrinsic dimension. Since the motion of hands, due to carrying of stuff or unwanted movements while speaking, is unpredictable compared to the movements of legs, in this paper, a binary mask is proposed for detachment of the down part of the body (legs) from silhouettes gathered from the considered video frames. The proposed method is applied to the Osaka database and shows an acceptable rate of identification compared to the case in which the whole silhouette of the image is considered.

Segregation of speech in presence of noise and other interferences is a challenging problem, especially in monaural case. Since accurate choice of time-frequency (T-F) units has great impact on segregation results, in this paper, we propose new methods for choosing T-F units in monaural voiced speech segregation. In the proposed system first, the units are marked using envelope cross-channel correlation in the initial segregation stage. Then, the system uses different features in the low and high frequencies for labeling of units in the pitch tracking and unit labeling stage. We use the enhanced envelope autocorrelation function (EEACF) for labeling T-F units in the high frequency range. We also consider the impact of each method on the segregation results. The evaluation of the monaural voiced speech segregation system based on different objective criteria isdone for different types of noise mixed with speech signals which shows better segregation results than conventional methods.

This paper presents an indirect adaptive type-2 fuzzy neural controller to control a class of nonlinear systems using feedback linearization theory and type-2 fuzzy neural networks. In this method, feedback linearization control input using two type-2 fuzzy neural networks is approximated. The gradient descent learning algorithm with adaptive learning rate is used for learning phase. In this paper, reduction in the number of fuzzy rules is proposed to simplify type-2 fuzzy neural network. By reducing the number of fuzzy rules, model complexity is diminished. Using Lyapunov stability theory, with the expression theorem, adaptation laws for tuning type-2 fuzzy neural networks is derived and stability of the closed loop system is guaranteed. Finally, the proposed controller is applied to twin-tank system in the presence of noisy data. Results are compared with Adaptive-Network-Based Fuzzy Inference System (ANFIS). Simulation results show the ability of the proposed method.