Catálogo de publicaciones - libros

Purpose of speech stream detection is to capture speech stream coming randomly in adverse acoustic environments. A novel robust method for speech stream detection is introduced based on both linear predict code all-pole model and lossless sound tube model to detect speech stream from inputs of wireless speech band communication. It makes use of autocorrelation distribution characteristics of variance sequence of linear predictive residual sequence to formulate two dimensions decision threshold vector. The decision threshold is adaptive to energy of background noise. It can make minimum decisions error. Plenty of signal stream data with various noises under various Signal-to-Noise Ratio and wireless speech band recordings on the spot were used to compare the proposed algorithm respectively with spectrum Entropy and short-time energy algorithm. The experiment results show that the new method for speech stream detection has good detection performance, and it performs well in adverse environments, and the speech stream detected sounds fluently.

In fast motion estimation, a search pattern with different shape or size has a very important impact on search speed and distortion performance. A motion estimation algorithm based on the novel large diamond and small pentagon search patterns is proposed in this paper. The stride of the proposed large diamond pattern is 3 pixels and it just need 2 or 3 search points for every new search step. So, the large diamond pattern can find the lager-motion vector quickly compare with the 2-pixel-stride hexagon pattern, and, it is does not easy to lose correct search path and fall into locally optimum point compare with the three-step search. The proposed small pentagon pattern can do more refined search than the small diamond pattern and small hexagon pattern. The proposed algorithm may find any motion vector regardless of no-, small-, medium- ,or large-motion with fewer search points than the diamond search algorithm and the hexagon-based algorithm while maintaining similar distortion performance. Experimental results substantially justify the further improvement achieved of the LDSPS algorithm compared with several other popular fast algorithms.

A shot boundary detection algorithm based on fuzzy theory and Adaboost is proposed in this paper. According to changes of color and camera motion, videos are classified into six types. By using features in compress domain such as DCT coefficients, the type of the MB, HSV color histogram difference, camera motion difference and so on, videos are segmented into three classes, that is, cut shot, gradual shot and non-change. The results of experiment have shown that this algorithm is robust for camera motion and walk-in of large objects in videos, and have better precision of shot boundary detection compared with the classic double-threshold method and the method of presented by Kuo et al .. There is no problem of threshold selection in our algorithm but it exists in most of other algorithms.

In this paper, it is reported that the method and primary application of a novel noninvasive technique, resting functional magnetic resonance imaging (fMRI) with unified statistical parameter mapping (SPM) independent component analysis (ICA), and principal component analysis( PCA), for localizing interictal epileptic activities of glioma foci. SPM is based on the general linear model (GLM). ICA combined PCA was firstly applied to fMRI datasets to disclose independent components, which is specified as the equivalent stimulus response patterns in the design matrix of a GLM. Then, parameters were estimated and regionally-specific statistical inferences were made about activations in the usual way. The validity is tested by simulation experiment. Finally, the fMRI data of two glioma patients is analyzed, whose results are consisting with the clinical estimate.

Design IIR digital filters with arbitrary specified frequency is a multi-parameter optimization problem. In this paper, we employ our proposed method, Quantum-behaved Particle Swarm Optimization (QPSO), to solve the IIR digital filters design problem. QPSO, which is inspired by the fundamental theory of Particle Swarm Optimization and quantum mechanics, is a global convergent stochastic searching technique. The merits of the proposed method such as global convergent, robustness and rapid convergence are demonstrated by the experiment results on the low-pass and band-pass IIR filters.

This paper considers the specific issues relating to finite word length (FWL) coefficient constraints for the case of infinite impulse response (IIR) digital filters. Due to the feedback nature of recursive filters, stability issues are an important factor in their design and are discussed in some detail. Some previously reported work on the optimization of FWL coefficients for IIR filters is also discussed. Extensive range of filter types and structures of IIR filters and their optimization using genetic algorithms is investigated and reported. Finally, comparative tests were conducted using the simple hill climber optimization techniques for a selection of filters.

The key frame technology is one of the most important technologies in the computer animation. All the researchers in the key frame algorithm are mainly focusing on the Euclidean object, without considering the fractal object. Fractal geometry can describe many irregularly shaped objects or spatially nonuniform phenomena in nature that cannot be accommodated by Euclidean geometry. In addition, the fractal objects in nature are almost colorful. Therefore, a computer aided inbetweening algorithm for color fractal graphics is presented in this paper. Our study shows that the fractal idea can be effectively applied in the key frame animation. The main feature of our algorithm is that it can deal with a fractal object while the conventional algorithm cannot. The theoretical and experimental results presented in this paper show that our algorithm has many practical values that can improve the efficiency of animation production and simultaneously greatly reduce the cost.

Feature sensitive hole filling is important for many computer graphics and geometric modeling applications. In this paper, we address the problem of reconstructing the salient features when filling holes in mesh. It respects fine shape features and works well on various types of shapes, including natural mesh and mechanical parts. For representing salient surface features, we adopt crest lines. The connectivity of the hole including the lost crest lines are decided by triangulation and region growing. The geometric position of the vertices and crest points in the hole are resolved by using a least-squares problem.

Speech signal detection is very important in many areas of speech signal process technology. In real environments, speech signal is usually corrupted by background noise, which greatly affects the performance of speech signal detection system. Correlation analysis is a waveform analysis method which is commonly used in time domain, and the similarity of two signals can be measured by using of the correlation function. This paper presents a new approach based on waveform track from cross correlation coefficients to detect speech signal in adverse acoustic environments. This approach firstly removes irrelevant signal so as to decrease the interference from noise by making use of computing cross correlation coefficients, and then decides whether contains speech signal or not according to the waveform track. Moreover, the performance of the algorithm is compared to the approach based on short-term energy and the approach based on spectrum-entropy in various noise conditions, and algorithm is quantified by using the probability of correct classification. The experiments show that the waveform from cross correlation coefficients is powerful in anti-interference, especially being robust to colored noise such as babble.

Contour information is always valuable for object analysis in image processing. In this paper, a new method of constructing contours of skin regions is proposed. To exploit skin formation in images, a nonlinear skin color classifier is first introduced. Then, a region splitting scheme is adopted to generate adaptive grids over skin regions. Based on the grids, initial contours are constructed. Finally, the contours are refined according to the minimum energy principle. Experimental results show that the proposed method has a good performance in contour construction of skin regions.