JOURNAL OF FACULTY OF ENGINEERING (UNIVERSITY OF TABRIZ)
Year:2006 | Volume:33 | Issue:1 (43) ELECTRICAL ENGINEERING|Papers Count:7

In this paper the characteristic loci features are used for the description of printed Farsi subwords. In extraction of these features, the numbers of crossings with subword bodies are restricted to 2. Using PCA, 12 uncorrelated features are selected. The images of subwords are clusterd using k-means algorithm with Euclidian distance. 9445 subwords of Lotus 12 font, with 400dpi resolution, are clustered to 150 and 300 clusters. Minimum and maximum cluster sizes are 11 and 91 subwords, respectively, for 150 clusters and 2 and 58 subwords, for 300 clusters. In a test, for clustering verification, images of 200 subwords, were rescanned and classified to 300 clusters. In this classification, the Euclidian distance from cluster means is used. In first, first five and first ten choices, 80.69%, 97.52% and 100% of these subwords were correctly classified.

This paper presents a new analytical method based on Genetic Algorithms (GA) for assessing the composite power system annualized reliability indices. In this proposed method, GA intelligently searches the numerous state spaces of a power system to find the most probable states contributing to system failure. A dynamic optimization load flow model is used for evaluation of sampled states. The proposed method is applied to a 6 bus RBTS and 24 bus RTS-96 reliability test system and the obtained result of program are compared with other conventional methods. Eventually the full set of composite system adequacy indices and load bus indices is calculated. The obtained results clearly demonstrate the capability of the proposed technique.

In this Paper a facial expression recognition system based on facial characteristic points and facial features extracted from them in frontal image sequences is presented. In the first digitized frame, 21 key feature points were manually marked with a computer-mouse around facial landmarks. Selected facial feature points were automatically tracked using a cross-correlation based optical flow. At first a Radial Basis Function (RBF) neural network were used to classify four facial expressions using feature vector extracted from position of Facial Feature Points (FFP) in the first and the last frame. Then 7 features were extracted from FFP's movement and were used to classify six basic facial emotions using RBF neural networks and Fuzzy Inference System (FIS). The recognition results from proposed classifiers showed that comparing with static image classifier, using image sequences gives higher recognition rate. Comparing RBF neural networks and FIS classifier systems shows that RBF classifier acts better than FIS classifier and gives higher recognition rate but it takes more processing time and for the real time implementation FIS classifier is preferred.

In this paper a Linearized Small- Signal Model for a six-pulse HVDC system is presented. In this paper the effects of on and off states resistances, forward voltage drop, filter and snubber circuits of the thyristors are taken into account. RL series circuits are used for modeling of the converter transformer and AC system. To validate the model developed in this paper, a comparison is made with the PSCAD/EMTDC time domain simulation program.

In this paper a new method for calculation of the harmonics in the AC current and DC voltage of a 6-pulse HVDC converter is presented. Instead of using FFT, the Fourier series is used for calculating harmonics. The HVDC converter is represented by a linearized small- signal model in mixed time- frequency domain which has been presented in part one of this paper [in this issue]. To validate the method, a comparison is made with PSCAD/EMTDC time domain simulation program.

Induction motors are the type of equipment most sensitive to voltage changes. In this paper behavior of induction motor fed from voltage source inverter under balanced and unbalanced voltage sags has been analyzed. Effects of various faults on the DC bus for six and twelve pulse rectifiers and induction motor have been analyzed. Voltage tolerance curves of induction motor for various rating capacitor and for various fault due to amount of allowable slip changes have been obtained. Keeping the drive online would be the best way of mitigating the effect of sag on the mechanical load. Since 66% of the faults are related to single-phase and phase-to-phase, by adding relatively amount of DC bus capacitance, these faults for drives can be tolerated.

In this paper, an adaptive beamforming algorithm is introduced. In this algorithm 2M2+7M multiplications are needed in every vector sample which M is the number of the array elements. It is possible to reduce the number of the multiplications to 5M if more simplifications are done which is suitable for practical applications. In this method, adaptive algorithm is applied in pattern space and the ratio of the main beam to the interference beam greater than 50dB is obtained which is acceptable in the considered environmentwith 6dB of SINR.