THE CSI JOURNAL ON COMPUTER SCIENCE AND ENGINEERING
Year:2005 | Volume:3 | Issue:1 (a)|Papers Count:5

The watershed transform is a conventional tool for the segmentation of images. Watershed segmentation is often not effective for textured image regions that are perceptually homogeneous. In this paper we describe a new image segmentation algorithm that integrates the measure of spatial variations in texture with the intensity gradients and consists of a number of conceptual stages. In the first stage, texture representation is calculated using vector summation of complex cell responses in different preferred orientations. In the second stage, gradient images are computed for each of the texture features, as well as for grey scale intensity. These gradients are efficiently estimated using a new proposed algorithm based on a hypothesis model of the human visual system. After that, combining these gradient images, a region gradient which highlights the region boundaries is obtained. Watershed transform of the region gradients properly segment the identified regions. Adaptive thresholding on rotational texture features is used to the problem of over segmentation. The combined algorithm produces effective texture and intensity based segmentation for natural and textured images.      

Packet classification is a central function for a number of network applications, such as routing, Quality of Service (QoS) provisioning and policy-based firewall deployment. A packet classifier categorizes incoming packets into specific flows, service classes or aggregates according to predefined rules. All packets belonging to the same flow obey a pre-defined rule and are processed in a similar manner by the router. Classification may, in general, be based on an arbitrary number of fields in the packet header. Performing classification on an arbitrary number of fields is known to be difficult, and has poor worst-case performance. Although there are significant previous works in this area, existing algorithms for packet classification reported in the literature scale poorly in either time or space as classifier grows in size. In this paper we propose a new method called Prefix-Based Bitmap Intersection (PBBI) which exploits some characteristics of actual classification rules to reduce memory consumption and classification time of previous methods. Mathematical evaluation and experimental results indicate a considerable performance improvement. It has been shown that our approach consumes memory less than previous work and has lower classification time. This makes our method suitable for implementing actual classifier even when the rule database grows largely.      

Increasing the domain of locality by using tree-adjoining-grammars (TAG) encouraged some researchers to use TAGs in applications such as machine translation, especially in the disambiguation process. Successful experiments of applying a TAG to French-English and Korean- English translation encouraged us to use it for another language pairs with very divergent properties, i.e., English and Persian. By using a Synchronous TAG (S-TAG) for this pair of languages, we can benefit from syntactic and semantic features of these languages. In this paper, we report on our successful experiments of automatic translation of English into Persian. We also present a computational model for disambiguation of lexical selection, based on a decision tree approach. Finally, a new automatic method for learning a decision tree from a sample data set is introduced    

Gaussian Mixture Models (GMM) and Support Vector Machines (SVM) exhibit uncorrelated error regions, so they can be combined to construct a classifier with higher performance. In this paper, an open-set text-independent speaker identification system is presented. In this system GMM capability of speaker modeling and discriminative power of SVM, are exploited in order to increase speaker identification accuracy. In training phase, using a validation database and GMM models, a confusion matrix is obtained. This matrix specifies groups of similar speakers. SVM models are trained to distinguish between speakers in each group. In identification phase, speakers are firstly identified by a first level GMM classifier. If the identified speaker falls in a group with similar speakers (confused speakers), second level classifiers i.e. SVMs are used to distinguish between speakers of this group. Identification error rate was reduced from 4.15%, when only GMMs were used, to 1.7% when identification was down by the proposed serial hybrid of GMMs and SVMs. Grouping of speakers was exploited to improve identification speed. In identification phase, first, the group of speaker is determined and then the speaker is identified in this group. This approach improved the identification time from 1.47s(in the case of base system using GMM sand no grouping of speakers) to 0.75s in the best case and 1.15s in the worst case. World model and maximum score normalization methods were also applied and evaluated for open set speaker identification. Both normalization techniques showed a considerable improvement in identification performance.      

In this paper after studying and analyzing BiCMOS and CMOS logical gates, we have achieved some advanced BiCMOS circuits for the same logical gates. In all of our designs, we used threshold detectors and voltage to current converters. In this new method, some parts of the conventional BiCMOS current-mode circuits have been eliminated without any changes in the logical functionality of the circuits, and some optimum circuits based on this logic family have been introduced. In the proposed circuits, we have obtained considerable speed gain and a remarkable reduction in transistor count.