فصلنامه پردازش علائم و داده ها
سال یازدهم شماره 1 (پیاپی 21، بهار و تابستان 1393)

In this paper a novel night sky star pattern recognition and precise centroiding approaches are proposed. Precision and computation time of image processing algorithm paly a great role in spacecraft in which the night sky star images are utilized for attitude determination. Star pattern recognition and centroiding are the most important steps of image processing algorithm in such attitude determination techniques. Here, in order to improve the computation time and precision of the image processing algorithm, a novel star pattern recognition approach including thresholding and clustering steps and precise centroiding method are proposed. Implementation results indicate that the proposed thresholding approach performs better than traditional approaches in dealing with images with uneven illumination. Lower computational burden and average centroiding error of less than 0.045 pixel obtained from experimenting 100 test simulated images, show the great capability of proposed image processing algorithm.

The stochastic active contour scheme (STACS) is a well-known and frequently-used approach for segmentation of the endocardium boundary in cardiac magnetic resonance (CMR) images. However, it suffers significant difficulties with image inhomogeneity due to using a region-based term based on the global Gaussian probability density functions of the inner\outer regions of the active contour. On the other hand, the local binary fitting (LBF) provides suitable results for segmentation of inhomogeneous regions because of employing a local Gaussian kernel function. In this paper, we propose a new active contour for inhomogeneous CMR images segmentation by substituting the region-based term of STACS with the corresponding energy functional of LBF. Furthermore, we automatically adjust weighting coefficients of the proposed energy functional according to the simulated annealing algorithm. The performance of the proposed method has been demonstrated on fourteen CMR images. All benchmark images are selected at the end of diastolic\systolic phase of the cardiac cycle. Furthermore, for each benchmark CMR image, the desired boundary was delineated by an expert. Experimental results demonstrated that compared to the geometric active contour, active contour without edge and STACS; the proposed method provides significantly superior performance for segmentation of the endocardium boundary of left ventricle of the human heart.

The rapid growth of published documents on the web has created some new requests for processing, classification and information retrieval. So, the use of natural language processing tools has increased around the world. Automatic summarization known as the core of a wide range of text-processing tools such as decision systems, accountability systems, search engines, etc. And always has been investigated as an important issue in computer science.This paper has introduced "Ijaz", a text summarization system, for Persian documents. For this, we first review the related works in this field, especially for Persian text summarization. We then investigate the using of some new effective features for improvement of the proposed summarizer system. Also for the first time, by using of a large corpus and standardized assessment tools, the proposed method has been evaluated and compared with other existing approaches for Persian text. The results of this evaluations are remarkable.

This paper shows how we can make advantage of using genetic programming in selection of suitable features for automatic modulation recognition. Automatic modulation recognition is one of the essential components of modern receivers. In this regard, selection of suitable features may significantly affect the performance of the process. In this research we implemented our model by using appropriate software and hardware platforms. Simulations were conducted with 5db and 10db SNRs. We generated test and training data from real ones recorded in an actual communication system. For performance analysis of the proposed method a set of experiments were conducted considering signals with 2ASK, 4ASK, 2PSK, 4PSK, 2FSK and 4FSK modulations. The results show that the selected features by the suggested model improve the performance of automatic modulation recognition considerably. During our experiments we also reached the optimum values and forms for mutation and crossover ratio, elitism policy, fitness function as well as other parameters for the proposed model.

3D model segmentation has an important role in 3D model processing programs such as retrieval, compression and watermarking. In this paper, a new 3D model segmentation algorithm is proposed. Cognitive science research introduces 3D object decomposition as a way of object analysis and detection with human. There are two general types of segments which are obtained from decomposition based on this principle: a core and salient parts. In this approach we start with calculating center of the model. Then, a point with maximum Euclidean distance from the center which represents a prominent part is chosen as the first salient point and its geodesic neighborhood points are deleted from salient point’s search domain. This process is continued until all salient points are detected. Then, the core part which connects the other parts to each other is detected. Thus, 3D model segmentation is completed. Considering center of the model as the reference point and utilizing both Euclidean and geodesic distance and deleting salient point’s neighborhood from salient point’s search domain led our proposed approach to be invariant against translation, rotation and pose changes and also decrease operation time of the proposed algorithm in comparison with the other 3D model segmentation algorithms.

Named Entity Recognition and Extraction are very important tasks for discovering proper names including persons, locations, date, and time, inside electronic textual resources. Accurate named entity recognition system is an essential utility to resolve fundamental problems in question answering systems, summary extraction, information retrieval and extraction, machine translation, video interpretation and semantic query expansion. Furthermore, named entity recognition can help us in some state-of-art problems such as removing ambiguity between two common names in different fields, finding out citations in scientific articles, recognizing the associations among persons and improving the results of a search engine to search queries containing named entities.Recently, many researches have been done on named entity recognition for English and other European languages which have led to efficient results; whereas the results are not convincing in Arabic, Persian and many of South Asian languages. One of the most necessary and problematic sub-tasks of named entity recognition is the person named extraction. In this article we have introduced a system for person named extraction in Arabic religious texts using "Proper Name candidate injection" by means of Conditional Random Field (CRF) method. Additionally, we have constructed a new corpus from traditional Arabic religious texts. Applying this method, our experiments have significantly achieved more efficient results.

In this study the distribution of sonority in syllable structure of Persian language is investigated. According to Sonority Sequencing Principle (SSP) sonority is minimum at the Onset, increases to maximum at the nucleus and decreases to last consonant of Coda. The results show that with some exceptions Persian language generally obeys SSP. The results also show that sonorant consonants are occurred more than they were expected no matter where their positions where. Non-sonorants such as plosives occur as expected in pre-vocal contexts which supports the idea that Perceptual Cue Salience plays an important role in shaping sonority sequencing in syllables. The results support the Licensing by Cue hypothesis.

This paper examines the performance of facial expression recognition improved by a discriminant analysis method. The proposed method incorporate knowledge of face clustering in linear discriminant analysis to improve it’s performance. Three face clustering approach have been considered. Intra-subject covariance and inter-subject covariance matrix are calculated in each cluster. The new sample is projected by mapping learned by samples in his cluster. Therefore, variation between train and test samples is reduced and the generalization performance of linear discriminant analysis is improved. Experimental results on the CK+ database confirm the efficiency of the proposed method in recognition rate improvement. This approach is applicable for many methods used in large scale facial expression recognition.

Text recognition has been one of the growing research topics in recent years. Many of these researches have focused on recognition of letters and sub-words as a basis for identifying larger text structures such as words, phrases and sentences. This thesis presents a new method in which the recognized sub-words are combined in order to provide meaningful words and sentences in Farsi texts. Since there may be more than one meaningful combination, the potential meaningful sentences are filtered using Farsi grammatical rules. In the sub-word recognition stage, a double scan method is exploited while the words are extracted using a database of frequent Farsi words. In the last stage a 2 and 3-gram method as well as Farsi grammatical rules are employed to identify the most meaningful sentence from all potential candidates. Experiments have proved the accuracy of the exploited method to be more than 85 percent.