THE CSI JOURNAL ON COMPUTER SCIENCE AND ENGINEERING
Year:2008 | Volume:6 | Issue:3 (A)|Papers Count:7

Wavelet-based denoinsing methods can be divided into thresholding and filtering based methods. The thresholding methods, which are based on the concentration of signal energy in a few number of wavelet coefficients, usually remove weak speech components. This tends to poor speech quality after removing colored noises due to speech distortion. Filtering based methods apply a filter to wavelet coefficients in order to denoise them and produce less speech distortion. In this paper, we first show the superiority of filtering methods in comparison to thresholding methods. For this purpose, we compare SURE soft thresholding method with Wiener filter. Then, we propose two new filters based on a posteriori distribution of wavelet coefficients. These new filters are compared to Wiener filter in the wavelet domain. Results show better performance of proposed filters. We also compare our filters with a maximum a posteriori filter in the frequency domain. Experimental results show that proposed filters improve signal to noise ratio and give a less distorted enhanced speech signal.

This paper presents a new model for an automatic intelligent traffic light system to measure traffic intersection from video images and controls the light timing program with calculating the occupation of inputs and outputs roads as a reliable sensor. It is based on background image deferential and updating that with three stages as creating real time background images, occupation calculation with eliminating shadows and traffic lights phases timings with high accuracy by a new method. System plays real time behaviors with accepting knowledge from the traffic knowledge based section and movement recalculating with commands generating. Control model of lights and phases is based on metapetrinet with timing depend on total capacity for all inputs and outputs camera's images of the four phase intersection. Automatically moving objects in another road sides against the movement direction of each road isn't effected on the results. The system processes 10 frames in 1 second for each camera. It is fault tolerant in congestion and shadow situations in smooth or suddenly light environments changes.

This paper presents a new method for evaluation of speech fluency. Speech fluency evaluation is needed in computer-assisted foreign language learning, where the fluency of speech uttered by a new language learner should be evaluated. Fluency evaluation helps the language learner to know his/her degree of speech fluency, so he/she can do more exercise and try to ameliorate his pronunciation fluency. To do this some quantitative measures are presented and a method has been proposed to automatically evaluate speech fluency using the best measures. In order to study the precision of the evaluation of the proposed method, a database of speech data from Turksish speakers was recorded and used as the benchmark in our studies. The correlation between the scores presented by human judges and the proposed automatic technique for measuring the fluency of uttered speech is about 95% which shows that the proposed technique performs as similar as the human judges. The proposed technique can be applied to other languages.

Ontology population refers to extracting the instances of classes (concepts) of an ontology from text. On the other hand Semantic Web provides structured content by adding annotations to documents. Semantic annotation tags class instances and map them into classes in an ontology. Thus annotation and population are two sides of a coin. This paper introduces an inter-language population and annotation system called POPTA. It annotates Persian texts according to an English ontology and populates the English ontology by instances extracted from Persian texts. POPTA exploits a combination of statistical and pattern based approaches besides web based methods, search engines, online encyclopedia and Wikipedia to perform the annotation and population tasks.

The Steiner tree problem in a weighted graph is to find a least cost sub tree in the graph such that it contains special vertices called terminal points. The Steiner tree problem on graphs is one of the best-known NP-hard problems. For this reason, many approximation algorithms for solving steiner tree problem have been proposed. The traditional algorithms achieve near optimal results, but cost more computational time. In this paper we propose an iterative algorithm based on learning automata for solving dynamic Steiner tree problem. Simulation results show that our proposed algorithm generates Steiner trees of low cost in reasonable running time.

Reducing the number of features is essential to improve the accuracy, efficiency and scalability of a classification process. There are two main reasons to keep the dimensionality of the input features: computational cost and classification accuracy. Reducing the number of input features can be done by selecting relevant features (i.e., feature selection) or extracting new features containing maximal information about the class label from the original ones (i.e., feature extraction). In this work, we use a mutual information based feature extraction (MIFX) algorithm for classification of electroencephalogram (EEG) in brain-computer interface (BCI). The tasks to be discriminated are the imaginative hand movement and the resting state. The experiments were conducted on four healthy subjects on different days. The results show that the classification accuracy obtained by MIFX is higher than that achieved by full feature set. Moreover, the results indicate that the performance obtained using MIFX is higher than that obtained using principle component analysis.

In the process of software architecture design, different decisions are made that have system-wide impact. One important decision is the selection of a suitable architecture style. The main problem in using architecture styles is the lack of investigation of their quantitative effect on software quality attributes. Concerning the vital role of reliability in safety critical software, one of the main criteria in choosing architecture styles of these systems is high reliability. The purpose of this study is to quantify the impact of architecture styles on software reliability that is a desired quality of safety critical software. We evaluate styles through two approaches of Reliability Block Diagram (RBD) and Markov State-based Model, and quantify their impact on software reliability. Then the results of two approaches will be compared. The main contribution of this paper is quantification and ranking of software architecture styles from the perspective of reliability quality attribute at stage of architectural style selection.