1 2538-4201 Research Center on Developing Advanced Technologies 714 Paper Pars Morph: A Persian Morphological Analyzer Mavaji Vahid Vazirnezhad Bahram 1 9 2011 8 1 3 8 22 09 2011 19 02 2018 In this paper, the theoretical foundation, the way of implementation and the uses of Pars Morph, a Persian morphological analyzer is introduced. Pars Morph is a rule-based Persian morphological analysis system, which analyzes the internal structure of word in Persian and determines the grammatical category and function of the word parts. Pars Morph being in link with a lexicon covering about 45000 lexemes and based on morphological rules in Persian can analyze the structure of complex lexemes and their possible inflected forms and even out-of-lexicon words. Accuracy rate of the first version of Pars Morph is about 95% which could be enhanced to 100% by applying the regulations on Persian language homographs, some syntactic features and typical features of Persian script. Pars Morph can be used in pure linguistics studies as well as in automatic analysis of Persian language.
708 Paper The role of Time Conjunctions in the Identification of Temporal Relations between Tensed-Verb Events in the Contemporary Persian Corpus 1 9 2011 8 1 9 16 22 09 2011 19 02 2018 This paper involves in prediction of temporal relation between tensed-verb events on the basis of conjunctions in texts. For this purpose, tensed verb event data were extracted from Contemporary Persian Corpus and were examined carefully. The temporal relations between events were identified. After analyzing data on the basis of temporal relation according to Bird’s and Allen’s categorization, it was revealed that temporal conjunctions can be categorized in 4 categories: determined, biased, pseudo random and random. Determined signals can predicate the temporal relation between the events accurately. Biased signals predicate one set of relations with a good estimation, Pseudo random signals doesn’t show any orientation toward any kinds of temporal relation and random signals make almost all kinds of relations possible. The results show that Bird’s prediction on the basis of signals is more desirable than Allen’s.   713 Paper Using Persian Stemmer in Information Retrieval System Faili Heshaam 1 9 2011 8 1 17 24 22 09 2011 19 02 2018 Using the language-specific behavior in information retrieval systems can improve the quality of the retrieved results significantly. Part of the word that remains after removing its affixes is called stem. Stemming process can be used for improving the relevancy of the results in information retrieval system. Different morphological variants of words (plural, past tense…) will be mapped into their stem which can be used in the searching process of information retrieval tasks. Using the stem instead of the surface of the word reduces the size of the index file significantly. In this paper, an algorithm for stemming Persian words is described and its effect on information retrieval system is evaluated with different ranking methods. By using Persian Porter stemmer with just 43 rules, the size of index file reduced about 5% while the mean average precision of the retrieval information system improved about 5%. 711 Paper Image Enhancement Using Gamma Correction 1 9 2011 8 1 25 32 22 09 2011 19 02 2018 In this paper a new automatic method is presented to enhance the image brightness through gamma correction process. Most of current gamma correction methods apply a uniform gamma correction across the image. Considering the fact that gamma variation for a single image is actually nonlinear, the proposed method does the gamma correction in a local approach. Thus the method is able to estimate appropriate gamma values for different regions of the image using a neural network. After windowing several training images with known gamma values, the mean and texture of each window (responsible for brightness and contrast of the window) are computed to train the neural network. The same features will be extracted from the unknown image to estimate the correct gamma values of the different parts of the image. Unlike other gamma correction methods, the proposed method does not change the gamma values of an image which does not need any brightness enhancement. The experimental results prove its better performance over other gamma correction methods. 709 Paper Improving face recognition from a single image per person via virtual images produced by imagination using neural networks abdolali fatemeh dadashi neda SeyyedSalehi SeyyedAli 1 9 2011 8 1 33 44 22 09 2011 19 02 2018 This paper deals with the problem of face recognition from a single image per person by producing virtual images using neural networks. To this aim, the person and variation information are separated and the associated manifolds are estimated using a nonlinear neural information processing model. For increasing the number of training samples in neural classifier, virtual images are produced for the neutral pose samples in a gallery dataset. By designing various structures of neural networks, the quality of virtually produced images, and consequently the recognition accuracy rate are improved. To obtain person information manifold codes giving better performance in describing the other persons and in generalizing, a learning method based on unsupervised clustering is presented. Applying this learning method and training classifier with virtual images, gives an accuracy rate of 83.63% on test dataset, which shows 12.73% improvement in comparison with training classifier using neutral pose samples. 707 Paper Application of Blind Source Separation for Speech-Music Separation Aghabozorgi sahaf Masoud Reza pishravian arash abutalebi hamid reza 1 9 2011 8 1 45 54 22 09 2011 19 02 2018 In this paper, the application of the Independent Component Analysis In this paper, the application of the Independent Component Analysis technique in speech-music separation is discussed. The separation algorithm is in the time domain. It needs the score function estimation to minimize the mutual information. For estimating score function, sufficient samples of the mixed (speech-music) signals are needed. In other words, these samples must be included both original sources. Since the speech and music signals could contain the silent gaps, the frame selection is important in our problem. Our proposed method for selecting the optimum frame is based on the score function difference. The experimental results show good performance of the proposed method in elimination of the silent gaps. Also they express the separation algorithm based on Gaussian Mixture estimator achieves a better separation performance and less processing time compared to the separation algorithm based on Minimum Mean Square Error estimator. 712 Paper Region-based Quality Improvement of Facial Images with Strong Shadows to Enhance Recognition 1 9 2011 8 1 55 66 22 09 2011 19 02 2018 Varying illuminations, especially the side lighting effects in face images, is one of the major obstacles in face recognition systems. Various methods have been presented for face recognition under different lighting conditions witch require previous knowledge about Light source and shadow area. In this paper, a novel approach based on H-minima transform to image segmentation and illumination normalization is proposed. Firstly, shadow area is extracted and modified by multi-stage method. Then, the gradient based criteria used to determine the best pattern of shadow areas. Subsequently, the obtained pattern of shadow areas is used to improve Retinex method in obscure area, obvious area, and all areas of the face image. Experimental results on the Extended Yale B database show that the proposed method significantly improves the performance of Retinex method. Furthermore, it provides the effective results in illumination normalization even in extreme lighting conditions.