SIGNAL AND DATA PROCESSING
Year:2013 | Volume:- | Issue:2 (SERIAL 20)|Papers Count:6

Atmosphere is a chaotic system and all climatic processes in this system have high order complexity. A massive volume of climatic data is collecting daily in synoptic stations of the country. This paper is a feasibility study of using these aboriginal data for short time weather forecasting. A framework is proposed that is based on a set of mappings from current weather state to 3 hours future state. In a part of paper, feasibility of 3 hourly forecasting for Hamedan based on current state of Hamedan and its 9 neighbor stations is studied. Mapping from current state to future state needs to selecting more important describetive parameters among all candidate parameters and using a regressioner for mapping to future state. Exprimental results show big gap between the precision of forecasts based on the data of synoptic stations and the precision of the forcasts produced by operational forecasting systems. Shortcomings of model may be one of the reasons of this gap, but it seems that the main reason is the lack of essential data in higher layers of atmosphere and low spatial resolution of used data.

Emotions play an important role in daily life of human, so the need and importance of automatic emotion recognition have grown with increasing role of Human Computer Interaction (HCI) applications. Since emotion recognition using EEG can show inner emotions, this method is more attention from other ways. In consideration to lack of emotion induction collection for doing such researches at Iranian culture, in this research some emotion induction experiments are designed to create four emotion states in subjects. Once subjects are experimented by International Affective Picture System (IAPS) that collected at Florida University and then they are experimented by corresponding movies with Iranian culture. Results show that corresponding movies get higher accuracy in comparison with IAPS. Fast computing, using only two electrodes and obtaining high accuracy from EEG signals are other advantages of this research.

In recent years, significant improvements have been achieved in statistical machine translation (SMT), but still even the best machine translation technology is far from replacing or even competing with human translators. Another way to increase the productivity of the translation process is computer-assisted translation (CAT) system. In a CAT system, the human translator begins to type the translation of a given source text; by typing each character the MT system interactively offers the choices to enhance and complete the translation. Human translator may continue typing or accept the whole completion or part of it. In this paper, we propose new approaches for increasing the performance of the interactive CAT. Our approaches are included a new search way and a hybrid back-off model. We could achieve 1.3% improvement by using our offered search approach.

In this paper for evaluating chaotic dynamics in improvement of robust pattern recognition in attractor recurrent neural network two models were proposed. The first model is designed based on natural selection; attractor recurrent neural network guides the evolution of chaotic nodes and these solutions intelligently. In second model, regarding to previous proposed models for chaotic neurons; the novel chaotic neural network was designed which the chaotic neurons have been placed in hidden layer of attractor recurrent neural network. By changing the parameters of chaotic neurons, their behaviors were changed in chaotic or periodic mode. The average performance of this model in recognizing noisy patterns in high level noise (more than 60%) is improved 8.5%, 37.16%, and 29.15% toward Attractor Recurrent Neural Network, Feedforward Neural Network, and Chaotic NDRAM respectively.Our proposed method improved the performance of first model, and attractor recurrent neural network 13.91% and 5.41% in high level noise respectively. Furthermore, this model has a better performance in comparison the attractor recurrent neural network also in non chaotic behavior (10.41% in high level noise).

Text tokenization is the process of tokenizing text to meaningful tokens such as words, phrases, sentences, etc. Tokenization of syntactical phrases named as chunking is an important preprocessing needed in many applications such as machine translation information retrieval, text to speech, etc. In this paper chunking of Farsi texts is done using statistical and learning methods and the grammatical characteristics of Farsi texts. Many features and labeling methods are examined one by one and the best features and labeling techniques are used for the detection of syntactic phrases and their boundaries. Several machine learning techniques including Support Vector Machine and Conditional Random Fields are used as classifier in our experiments. The impact of the size of training texts on chunking performance was studied as well. Using the proposed methods in this paper, a performance of 84.02% was obtained for detection of phrase boundaries and 78.04% for detection of both phrase boundaries and phrase type.

Video text detection plays an important role in applications such as semantic-based video analysis, text information retrieval, archiving and so on. In this paper, a Farsi/Arabic text detection approach is proposed. First, a three-level resolution pyramid of input image is created. Then, with an appropriate edge detector, edges are extracted and then by using edges cross points, artificial corners are extracted. Artificial corner histogram analysis is done for rejecting non-text corners. The discrete cosine transform (DCT) coefficients of input picture are extracted and texture intensity picture is created by combining appropriate coefficients. With combining artificial corners image and texture intensity image, a features vector is extracted and fed into support vector machine (SVM) classifier for detecting text regions. Finally, with drawing normalized texture intensity profiles, final verification is done and text lines are separated from each other’s.