Earticle

This paper presents a novel algorithm of speech enhancement using data adaptive soft-thresolding technique. The noisy speech signal is decomposed into a finite set of band limited signals called intrinsic mode functions (IMFs) using empirical mode decomposition (EMD). Each IMF is divided into fixed length subframes. On the basis of noise contamination, the subframes are classified into two groups – noise dominant and speech dominant. Only the noise dominant subframes are thresholded for noise suppression. A data adaptive threshold function is computed for individual IMF on the basis of its variance. We propose a function for optimum adaptation factor for adaptive thresholding which was previously prepared by the least squares method using the estimated input signal to noise ratio (SNR) and calculated adaptation factor to obtain maximum output SNR. Moreover, good efficiency of the algorithm is achieved by an appropriate subframe processing. After noise suppression, all the IMFs (including the residue) are summed up to reconstruct the enhanced speech signal. The experimental results illustrate that the proposed algorithm show a noticeable efficiency compared to the recently developed speech denoising methods.

Temperature modulation of metal oxide semiconductor (MOS) gas sensors has been widely used due to its higher discriminating power. The temperature modulation alters the kinetics of the gas-sensor interaction leading to characteristic response patterns. However, the selection of frequencies and duty cycles is based on trial and error method. In this paper, we have introduced a method to systematically determine the optimal set of modulation frequencies and duty cycles using system identification theory for sensor modeling. Pulse modulation being a popular method of feature extraction of MOS sensors, optimization of parameters of pulse modulation becomes very significant. In our work, system identification has been applied to select the sensor model that provides the most stable and desired sensor response, hence solving problem of choosing the best frequency and duty cycle of the temperature modulating signal of the MOS sensor. The estimation of model parameters is done using iterative prediction-error minimization (PEM) method. The best suited transfer function was chosen for the MOS gas sensors based on the sensor stability and then the sensors were operated at the respective best frequencies and duty cycles. Principal Component Analysis (PCA) was used to visualize the different sample gas patterns. Data classification was performed using supervised neural network classifiers; namely the Multi-Layer Perceptron (MLP) network and Radial Basis Function (RBF) network and the classification percentage before and after optimization were compared henceforth.

This paper presents the evaluation of two approaches, widely used in the inpainting literature, applied in the context of the atmospheric noise removal, such as fog, clouds - dense and sparse - and shadows, which often occur in remote sensing images. The presence of such elements aect, in many ways, the image processing in an environmental or urban monitoring, and also in the steps of the digital image processing, suchlike segmentation and classication. Whilst one approach uses a technique of interpolation for the dissemination of information by a multidimensional Discrete Cosine Transform (DCT) smoothing method, the other one is based on second-order partial dier- ential equations methods (PDE). This PDE-based development uses the heat diusion and thin-plate spline methods to achieve their solutions with the aid of the nite-dierence method. To proceed the methods evaluation, this work uses the kappa coecient and the Peak Signal-to-Noise Ratio index (PSNR). The metrics indicate the eectiveness of the DCT strategy, which produces higher quality images, specially when comparing the results obtained by the use of dierential equations modeled by thin-plate spline. The visual aspect of images is clearly an important factor for measuring the eectiveness of any image processing, so, in addition to numerical metrics, are presented the nine images used to evaluate both methods.

One of important components in an image retrieval system is selecting a distance measure to compute rank between two objects. In this paper, several distance measures were researched to implement a foliage plant retrieval system. Sixty kinds of foliage plants with various leaf color and shape were used to test the performance of 7 different kinds of distance measures: city block distance, Euclidean distance, Canberra distance, Bray-Curtis distance, 2 statistics, Jensen Shannon divergence and Kullback Leibler divergence. The results show that city block and Euclidean distance measures gave the best performance among the others.

Noise is the major difficulty that arises in echocardiographic image processing. The problem is especially important if the noise has a multiplicative nature (speckle noise).In this Paper we introduced a hybrid filter. This filter is combination of Speckle removal SRAD and Relaxed median filter. Experiments are carried out on Echocardiographic images. Results show Speckle reduced edge preserved images. This hybrid filter removes speckle noise as well as removes impulse noise. This model also enhances the edges in the image.

This paper presents some image processing techniques that can be used for radiographic image enhancement. Contrast enhancement, filtering, denoising, and interpolation processes are carried out in this paper. Contrast enhancement is carried out using adaptive histogram equalization. Filtering is carried out using median, Wiener, Lee, and Kuan filters. Wavelet and curvelet transforms are used for image denoising. Three interpolation are carried out. The results are evaluated qualitatively and quantitatively using the Peak Signal-to-Noise Ratio (PSNR), Root Mean Squire Error (RMSE), Standard Deviation (SD), smoothness, entropy, Structural Similarity (SSIM), and execution time. The results show that the contrast enhancement improves the radiographic image quality, the Wiener filter achieves better enhancement results than other filters, the curvelet transform denoising gives better enhancement than wavelet denoising. The bicubic interpolation with resolution factor two is promising in terms of the quality assessment metrics.

Detection of cardiac arrhythmias, particularly ventricular fibrillation (VF), and ventricular tachycardia (VT) have been highly regarded and has done several works in this field. In this study, a method based on the Takagi-Sugeno-Kang (TSK) fuzzy system for ECG arrhythmia detection and classification of normal sinus rhythm (NSR), ventricular fibrillation (VF) and ventricular tachycardia (VT) has been used. ECG arrhythmia signals have been obtained from MIT-BIH database. At the first, preprocessing is performed on the signals to get a signal without any noise. Then two features of ECG signals include an average period T (i.e. the time interval between two R peaks) and amplitude of QRS complex, are used as inputs to fuzzy classifier. The triangular membership functions for converts crisp input values (features of ECG signals) to the fuzzy values are used to provide the fuzzy system. Using genetic algorithms, optimization rules with membership functions by minimizing the error function, and convert them to proper rules and membership functions for classifying arrhythmias do with high accuracy. Finally, we achieved the classification accuracy for normal signals (NSR) 91.66%, for VT signals 92.86% and for VT signals equal to 100%. We obtained the overall accuracy of the classifier 93.33%. Also, sensitivity for NSR signals is equal 92.30%, for VT signals is 93.33% and for VF signals is equal to 100%. Specificity for NSR, VT and VF signals is equal to 94.44%, 93.57% and 100% respectively. The simply of propose method can be considered as its major advantage.

The performance of a number of image processing methods depends on the output quality of a thresholding process. Typical thresholding methods are based on partitioning pixels in an image into two clusters. In this paper, a new thresholding method is presented. The main contribution of the proposed approach is the application of the empirical mode decomposition (EMD) on detecting an optimal threshold for an input image. The EMD algorithm can decompose any nonlinear and non-stationary data into a number of intrinsic mode functions (IMFs). When the image is decomposed by empirical mode decomposition (EMD), the intermediate IMFs of the image histogram have very good characteristics on image thresholding. The experimental results are provided to show the effectiveness of the proposed threshold selection method.

Segmentation of pulmonary X-ray computed tomography (CT) images is a precursor to most pulmonary image analysis applications. Digital Image Processing is currently a hot research area in medicine and it is believed that they will receive extensive application to biomedical systems in the next few years. In Digital Image Processing, neural networks are ideal in recognizing diseases using scans since there is no need to provide a specific algorithm on how to identify the disease. This paper describes an algorithm to separate the lung tissue from a Chest CT to reduce the amount of data that needs to be analyzed. Our goal is to have a fully automatic algorithm for segmenting the lung tissue, and to separate the two lung sides as well. Fuzzy c-Means clustering is used to segment the lungs. Cleaning is performed to remove air, noise and airways. Finally, a sequence of morphological operations is used to smooth the irregular boundary. The database used for evaluation is taken from a radiology-teaching file. Our current evaluation shows that the applied segmentation algorithm works on a large number of different cases. The textural features were extracted from the segmented lungs and it was given as input to CFBP. The neural networks are used to identify the various lung diseases.

In this paper, we propose a method for character type classification based on a probabilistic topic model. The topic model is originally developed for topic discovery in text analysis using bag-of-words representation. Recent studies have shown the model is also useful for image analysis. We adopt the probabilistic topic model for character type classification. In our method, character type classification is carried out by classifying image patches based on their topic proportions. Since the performance of the method depends on a visual vocabulary generated by image feature extraction, we compare several feature extraction and description methods, and examine the relations to classification performance. In addition, by extending the method, we propose a coarse-to-fine approach to achieve stable character type classification for a small image patch. For that purpose, firstly, we partition an image into several patches which contain enough information to estimate the model parameters via EM algorithm. Then, each patch is subdivided into smaller patches. Estimation on the small patch is carried out by MAP-technique with a prior reflecting topic proportion of its parent patch. Through the experiments, we show accurate character type classification is made possible by the probabilistic topic model.

The detection of moving object is the hot topic of computer vision research. In this paper, we propose a new method to detect moving object in surveillance video based on the visual features of frog eyes. The method used the principle of entropy which derived from information theory. First, we filtering out the noise and some unnecessary information in the frame images sequence used smoothing filter; Second, detecting the edge point of the smoothed image obtained in the first step by Canny operator; At last, spatial-temporal sliding windows are built for each edge point, then calculating the entropy of spatial-temporal windows and comparing the value in adjacent frames, in which the difference exceeds a certain intensity is considered as a part of moving object. Experimental results show that the proposed method can detect the edge of moving object effectively and has lower computational complexity.

The target echo of downward-looking radar can be simulated by a multicomponent Chirp signal with noise, which makes signal detection and parameter estimation a very important processing. Aiming at the problems of error detection caused by the cross-terms in WVD-Hough methods and other modified WVD distribution methods, and low precision due to poor time-frequency concentration in S-Hough transform, this paper proposed a signal detection and parameters estimation algorithm of multicomponent Chirp signals based on generalized S-transform. Firstly calculate the time-frequency distribution of multicomponent Chirp signals by generalized S-transform, then execute Hough transform to the generalized S-transform TFD on the time-frequency plane to implement signal detection and parameter estimation to the multicomponent Chirp signals. The simulation results show that this algorithm has good performance in detection the multicomponent Chirp signals in low SNR case.