SIGNAL AND DATA PROCESSING
Year:2009 | Volume:- | Issue:2 (SERIAL 12)|Papers Count:7

Speech databases are part of the concatenative text-to-speech (TTS) systems. Quality of the databases and the way of their development play a significant role in the naturalness of the synthesized speech. This paper introduces a syllable speech database and a diphone speech database and describing their characteristics. The phases in their development are explained. Then differences and advantages of these two databases are discussed using Persian text-to-speech synthesis system “Gooya”.

Part-Of-Speech (POS) tagging is essential work for many models and methods in other areas in natural language processing such as machine translation, spell checker, text-to-speech, automatic speech recognition, etc. So far, high accurate POS taggers have been created in many languages. In this paper, we focus on POS tagging in the Persian language. Because of problems in Persian POS tagging, a comprehensive plan is proposed to reach a high efficient POS tagger in this language. Afterward, morphological analysis is investigated in Persian and it is shown that using a morphological analyzer in inflection level, POS tagging has been improved greatly. The results describe the fruitfulness of the proposed method.

Automatic classification of lung tissue patterns in high-resolution computed tomography (HRCT) images of patients affected by interstitial lung diseases (ILD) is an important stage in the construction of a computer-aided diagnosis system. In this study, we propose a new image based system for classification of lung tissue patterns. The proposed system comprises three stages. In the first stage, the parenchyma region in HRCT lung images is separated using a set of thresholding, filtering and morphological operators. In the second stage, two sets of over complete wavelet filters, namely discrete wavelet frames and rotated wavelet frames are utilized to extract features from defined regions of interest (ROIs) within parenchyma. Then, in the third stage, the fuzzy k-nearest neighbor algorithm is employed to perform the pattern classification. The proposed method is tested for classifying four different lung tissue patterns (ground glass, honey combing, reticular and normal) selected from a database of 340 images from 17 subjects. After applying our technique to classify these patterns in isolated ROIs, we extend the classification scheme to the whole lung in order to produce quantitative scores of abnormalities in lung parenchyma of patients. The performance of the proposed method is compared with two state-of-the-art texture based methods for lung tissue characterization and is also validated against experienced observers. The average kappa statistic of the agreement between two radiologists and the computer was found to be 0.6543 whereas the average kappa statistic for the inter-observer agreement was 0.6848. We also conducted an experiment to show the correlation between Pulmonary Function Test parameters and quantitative scores of computerized system. Results show that the extent of HRCT findings correlates significantly with functional impairment. The computer system is shown to approach the performance of the expert observers in diagnosing regions of interest and can help to produce objective measures of abnormal patterns in lung HRCT images.

The purpose of this study was to apply recurrence plots on event related potentials (ERPs) recorded during memory recognition tests. EEG signals recorded during memory retrieval in four scalp region were used. Two most important ERP’s components corresponding to memory retrieval, FN400 and LPC, were detected in recurrence plots computed for single-trial EEGs. In addition, the RQA was used to quantify changes in signal dynamic structure during memory retrieval, and measures of complexity as RQA variables were computed. Given the stimulus, amplitude of the RQA variables increases around 400ms, corresponding to dimension reduction of system. Furthermore, after 800ms these amplitudes decreased which can be as a consequence of an increase in system dimension and complexity and back to its basic state. The mean amplitude of Old items was more than New one. Furthermore we applied statistical analysis (t-test) to find meaningful difference between features extracted from nonlinear measures. Using this method, we found its ability to detect memory components of EEG signals and to do a distinction between Old/ New items. In contrast with linear techniques, recurrence plots and RQA do not need large number of recorded trials, and they can indicate changes in even single-trial EEGs. RQA can also show differences between old and new events in a memory process.

Medial entorhinal cortex is known to be the hub of a brain system for navigation and spatial representation. These cells increase firing frequency at multiple regions in the environment, arranged in regular triangular grids. Each cell has some properties including spacing, orientation, and phase shift of the nodes of its grid. Entorhinal cortex is commonly perceived to be the major input and output structure of hippocampal formation; grid cells are one synapse upstream of place cells in hippocampus. The problem is how single confined place fields can be generated from the repetitive activity of grid cells. In this article we have proposed an artificial neural network model based on radial basis function, which allows for the single confined place fields of hippocampal pyramidal cells to be emerged from the activities of grid cells. In order to evaluate the performance of the model, it was considered in two steps in a one-dimensional and two-dimensional environment. Simulations were done considering different characteristics of grid cells and the model demonstrated a good performance in generating single spot activity for place fields.

We consider a new scenario in blind speech separation problem in which the number and the features of active sources change with time in opposite to the previous methods in which all sources are active all the time. Accordingly, we propose the new DESICA algorithm for source separation which is a compound of the ICA and DESPRIT algorithms. In this algorithm, using the ICA, the separation process is performed initially and then using the DESPRIT algorithm, the binary mask of silence intervals is calculated. Finally, by applying the binary mask to the separated signals, the final separation is obtained. Simulation results show that the DESICA algorithm improves the SDR and SIR about 6dB compared to those of the DESPRIT algorithm. Also, the SEN is improved about 11 and 17dB with respect to those of the ICA and DESPRIT algorithms.

JPEG is the most applicable image format in digital communication. In recent years, various steganography methods have been proposed for it. This paper aims to study and classify JPEG steganography schemes and introduce different methods to improve their security based on cover. Accordingly, the effective factors in security that are related to the cover such as double compression, spatial frequencies and quality factor have been evaluated theoretically and experimentally. Also some well-known steganography algorithms and software tools have been introduced, evaluated and classified based on different criteria. Some of these algorithms have been implemented in Stegotest software and the destruction effects of steganography in different methods have been compared.