Earticle

Difference co-array can be constructed by using Khatri-Rao (KR) product, which can increase the degrees of freedom (DOF) significantly. Combined with fourth order cumulants, the KR product can be used to constructed fourth order difference co-array. The fourth order difference co-array of a four level nested array contains a uniform linear array (ULA), however, its second order difference co-array has missing holes, which may result in the ambiguity for DOA estimation. And the method based on KR product and fourth order cumulants has two main drawbacks. First it cannot be employed to Gaussian source signals. Second it needs a large number of snapshots. In this paper, a novel approach is proposed to construct a virtual ULA based on KR product and redundancy spacing of arrays for a four level nested array. Unlike the existing method based on KR product and fourth order cumulants, the new method only uses second order statistics. And compared to the method based on KR product and second order statistics, the new method achieves higher resolution. Numerical results are provided to demonstrate the effectiveness and superior performance of the proposed algorithm.

Sampling rate conversion is, in general, implemented using time domain based poly-phase filter structures. The other way of implementing multi-rate signal processing is frequency domain based approach, which has the advantage of computational savings. This paper clearly explains how to apply frequency domain processing techniques for sampling rate conversion. The proposed approach is applied to both power of 2 and non-power of 2 sampling rate conversion factors. The properties of FFT are utilized analytically to solve the implementation problems such as non-power of 2 sampling rate factors. The theoretical computational complexity of the proposed approach is provided. The simulation results of proposed approach are compared with the quality of time domain approach and the comparison shows that differences are insignificant.

Textural patterns can often be used to recognize familiar objects in an image or retrieve images with similar texture from a database. Texture patterns can provide significant and abundance of texture and shape information. One of the recent significant and important texture features called Texton represents the various patterns of image which is useful in texture analysis. However sometimes the textured image obtained may not be of good quality and this may lead to improper detection of significant patterns. To enhance the quality or better illumination or contrast or sharpening effect, the present paper applied various local grey to grey level preprocessing steps on textured data. Grey to grey level preprocessing is required because the patterns on textons can be evaluated only on grey level textures. The present paper evaluated the occurrence behavior of various texton patterns based on the various grey to grey level pre processing methods, for an efficient rotationally invariant texture classification. The experimental results on various stone textures indicate the efficacy of the proposed method when compared to other methods.

Coastal surveillance and harbour defence are the most complex and challenging opera- tional issues for modern navy in the current turbulent global political climate. In most of the coastal surveillance and harbour defence systems, long sea-bed arrays consisting of hundreds of pressure sensors are deployed along the coastal belt to capture the low frequency compo- nents emanating from the sub-surface targets. Deployment of these sensor-arrays along with its associated signal conditioning hardware at the ocean-bed is a challenging task. The output of the sensor-array is to be conditioned and then digitized using multi-bit analog to digital converters (ADC). Further, the digitized channel data are required to be send to a base station through a radio frequency link. In this paper, we propose a compressively sampled (CS) architecture of acoustic vector sensor (AVS) array, to estimate the direction of arrival (DoA) of multiple acoustic sources, in a range independent shallow ocean using a one-dimensional search without prior knowledge of the ranges and the depths of the sources. We extend the high resolution angular spectral estimators MUSIC, MVDR and subspace in- tersection method (SIM) to suit the compressively sampled AVS array architecture operating in a shallow ocean environment. This architecture promises a signicant reduction in the number of sensors, analog signal conditioning hardware, data rate or bandwidth, the number of snapshots and the software complexity, leading to easy installation and maintenance.

Shape is one of the main low level features in content based image retrieval systems (CBIR). This paper proposes a novel CBIR technique based on shape feature. In this technique feature extraction is based on a rectangle that covers a shape. The proposed signature is a Fourier based technique and it is invariant against translation, scaling and rotation. The retrieval performance between some commonly used Fourier based signatures and our Smallest Rectangle Centroid Distance (SRCD) signature has been tested using the MPEG-7 database. Experimental results show that the SRCD signature has a good performance compared with those shape signatures.

To suppress the Pseudo-Gibbs phenomena caused by the Contourlet, the Nonsubsampled Pyramids Filter Banks and the Nonsubsampled Directional Filter Banks are combined to construct the nonsubsampled Contourlet transform (NSCT). Hence, The NSCT not only possess the main features of multi-scale, multi-directional and time-frequency localization, but also offer the property of the shift-invariant which is vital to image processing. Firstly, multi-scale decomposition is performed on source images using NSCT to get high-frequency and low-frequency images. Secondly, the Novel Sum-Modified-Laplacian and Local Neighbour Sum of Laplacian are respectively used to select the lowpass coefficient and highpass coefficients to combine fused image. Finally, the inverse nonsubsampled contourlet transform is applied to obtain fused image. Experimental results show the proposed approach outperform the traditional discrete wavelet transform-based and the Contourlet-based image fusion methods.

Among the various hand gestures, pointing gesture is highly intuitive, and does not require any priori assumptions. A major problem for pointing gesture recognition is the difficulty of pointing fingertip tracking and the unreliability of the direction estimation. A novel real-time method is developed for pointing gesture recognition using Kinect based depth image and skeletal points tracking. An adaptive virtual touch screen is constructed instead of estimating pointing direction. When a user stands in a certain distance from a large screen to perform pointing behaviors, he interacts with the virtual touch screen as if it is just right in front of him. The proposed method is suitable for both large and small pointing gestures, and it’s not subject to users’ characteristics and environmental changes. Experiments have highlighted that the proposed approach is robust and efficient to realize human-computer interaction based on pointing gesture recognition by comparisons.

In this paper, we propose a Copyright Protection (CP) watermarking scheme, which is a robust, blind, and reversible watermarking scheme; aimed at protecting the copyright of color images against geometrical attacks. It is a development of an already existing watermarking scheme. Watermark embedding in this scheme, as well as in its counterpart, mainly depends on the permutation of the histogram bins. However, we present a new embedding rule, which increased the average capacity by about 55 bits, and it increased the quality, PSNR, of the watermarked image from 35.01 dB to 38.05 dB. The embedded watermark demonstrates 100 % robustness against a variety of geometrical attacks, like Flipping (H, V, and Both), Rotation (90°, 180°, and 270°), Scattering, Warping, Skewing, and their combinations. Finally, our proposed CP watermarking scheme showed a faster watermark embedding process than that of its counterpart by an average reduction in time equals, 4.84 seconds.

Since introduction, the Support Vector Machines (SVM) has been popularly used in machine learning and data mining tasks due to their strong mathematical background and promising result. Nevertheless, they are noticeably slow in the prediction stage. The speed is influenced by number of support vectors determined in the training phase. Motivated by this fact, several studies are done to reduce the number of support vectors. The reduction should consider the degeneration of learning quality and preserve it at much as possible. Most previous methodologies either reduce the training set or apply a post-processing step to reduce the number of support vectors. In this paper, we proposed a new SVM cost function called Step Regularized Support Vector Machine (SRSVM), which is a standard SVM with extra constrained to reduce the number of support vectors, which can be defined by user. Experimental results are done to evaluate the efficiency and speed of proposed algorithm. SRSVM are also compared to other related SVM algorithms. The comparisons showed that the proposed method is effective in reducing number of support vectors while preserving the high performance of the classifier.

In today’s world face recognition, applications attain great popularity due to best application of image analysis and availability of feasible technologies. It has a wide scope of applications in different fields like pattern recognition and commercial market. A large number of researches have been done research within this field. The different area's Researchers, including computer science and neuroscientist are working within this field. Due to distinct variation in facial expression, occlusion and illumination, face recognition process becomes more challenging from last few decades. In this paper, we critically evaluate different state-of-the-art face recognition techniques. Strengths and weakness of different techniques are identified, which helps in the future research.

A class of neural filter for image enhancement is proposed in this paper. The proposed intelligent filter is carried out in two stages. In first stage the corrupted image is filtered by applying two special classes of decision based filters. Filtered image outputs from decision based filters are suitably combined with a Feed forward neural network in the second stage. The internal parameters of the feed forward neural network are adaptively optimized by training for three well known images. This is quite effective in eliminating impulse noise. Extensive simulation results show that the proposed filter is superior in terms of eliminating impulse noise as well as preserving edges and the results are compared with other existing filters.

This study proposes a post-processor to reduce the effect of musical residual noise which is annoying to the human ear. Initially, a state-of-the-art speech enhancement algorithm is performed as the first stage to reduce background noise for noisy speech. Hence the enhanced speech is post-processed by a hybrid-mean filter to reduce the musical effect of residual noise. In the case of a vowel-like spectrum, directional-mean filtering is performed to slightly reduce the musical effect of residual noise, where the harmonic spectrum can be well maintained at an acceptable level. Conversely, block-mean filtering is performed to heavily reduce the spectral variation for noise-dominant spectra, enabling musical tones to be significantly smoothed. The musical effect of residual noise is therefore reduced. Finally, the pre-processed, the directional-mean filtered and the block-mean filtered spectra are fused according to speech-presence probability. Experimental results show that the proposed hybrid-mean filter can efficiently improve the performance of a speech enhancement system by reducing the musical effect of residual noise.

Biorthogonal wavelet transform has been widely used in the fields of image denoising and image coding. It is usually realized by linear phase filters. All phase digital filter is a new type of linear phase digital filter which has been proposed in recent years and has frequency structure. On the basis of analyzing principle of biorthogonal wavelet transform and discrete cosine sequency filter (DCSF), this paper proposes a new algorithm to implement biorthogonal wavelet transform by using discrete cosine sequency filter. Simulation experiments to typical test images are conducted in MATLAB. Experimental results show that whether using the discrete cosine sequency filter or the biorthogonal wavelet transform, the results of the decomposition and reconstruction are the same. It comes to conclude that the proposed algorithm is valid.

To solve the difficulty in recognition coal-rock interface for the top coal caving process, we proposed a new method based on Mel-frequency cepstrum coefficient (MFCC) and neural network. In this paper, we conducted the noise separation by Independent Component Analysis (ICA) for acoustic signal. Then we extracted MFCC as the feature and recognized the coal-rock interface via BP neural network. The result shows that MFCC reflect the voice features of coal-rock more effectively, comparing to other features (frame energy and kurtosis), it provides average relative reductions of 12% and 19% in error rate, which recognition rate is 83%. We conclude that the method based on MFCC and neural network is an effectively and automatically detection for the coal-rock interface recognitions.

Directly from the brain thinking activity signals to communicate with the outside world, to achieve the heart and heart communication, achieve control of the surrounding environment, even is the dream of human beings since the ancient times is the pursuit of. Brain-computer Interface (Brian - Computer Interface: BCI) this novel human-computer interaction mode provides the scientific way to realize this dream. People hope that the new communication technology can be used in traffic tools, weapons, and other auxiliary control system, especially for those neuromuscular damage, cannot use the conventional methods of communication disability patients provides another way to communicate with the outside world. Exercise imagination refers to through the brain consciously simulate a certain action, but without obvious physical activity. In the human brain has a corresponding motor cortex area, when people have limbs activities, the motor cortex area is active. In imagine movement, although physical activity, but has remained active in the areas of the brain's corresponding motor cortex, the brain also sends out the corresponding EEG signals, so that there will be movement similar brain electrical signal, but due to the body don't exercise, avoid the my electricity interference, using the movement of the thought mainly, participants imagine left and right hand movement, or don't want to, the need to constantly training, participants learn to imagine the essence of sport, to avoid other distractions. So-called brain-computer interface, it is an organization that does not depend on peripheral nerves and muscles, etc. Usually the brain output channel of communication system. In recent five years, the research of this field gradually formed a hotspot; dozens of research team in the world have developed various forms of BCI experiment system. This research mainly based on multiple electrodes EEG recording, for a variety of brain stimulation mode is intended to explore the spatial and temporal variations of electrical signals. Applied to the second-order blind identification, phase synchronization and energy entropy of the signal analysis methods to analyze imagine movement EEG signals processing, extracting its features, and USES the BP neural network and support vector machine (SVM) classification method for different types of EEG classification is imagine movement, won a higher classification accuracy and designed a BCI system based on motion imagination, through this system, participants can more freely to imagine to control the mouse movement or virtual car movement to the left or right. The innovation of this study is to imagine the movement of brain electrical signal as input signal of the brain-computer interface system, imagination is a very complicated process, and the brain electrical signal characteristic is not obvious, so higher requirements for feature extraction and classification algorithm.

This paper presents an accurate optical ow estimation approach by reformulating the conventional image brightness constraints and representing the ow in sphere coordinate. A more sophisticated representation for optical ow is derived to substitute current ow modeling. The new model provides better understanding of the behavior and limitations of conventional methods in terms of brightness constraint, and the sphere representation leads to modied algorithms that outperform traditional ow approaches. This seemingly small change in representation provides more direct access to the inner characteristics of a ow led. We compared to the conventional ow methods on standard data sets and show that we achieve more accurate and promising results.

The image segmentation result based on level set typically depends on the appropriate manual initial contour. In this paper, we introduce an autonomous approach for deciding the initial level set contour to be close to the actual boundary as far as possible, and the decided initial contour can be directly evolved by various level set methods. Such an improvement can speed up the evolution and lead to a more robust segmentation result. Then, we consider the statistical information of three distinct regions to construct a new level set model, including contour, contour inside and outside. Combining the two steps above is helpful to obtain a pretty ideal segmentation effect. Some remarkable results and shorter execution time for some difficult segmentation tasks shown in this paper demonstrate the potential of our innovative approach.

In practice, Financial Time Series have serious volatility cluster, that is large volatility tend to be concentrated in a certain period of time, and small volatility tend to be concentrated in another period of time. While GARCH models can well describe the dynamic changes of the volatility of financial time series, and capture the cluster and heteroscedasticity phenomena. At the beginning of this paper, the definitions and basic theories of GARCH(1,1) models are discussed. Secondly, show the heavy tail behavior of GARCH(1,1) process with α-stable residuals {}ttZε∈, (0,2]α∈ and {}ttZε∈ errors. In fact, both these processes have heavy-tailed properties, but generally the tail of GARCH(1,1) process is heavier than the tail of {}ttZε∈ errors. And then the modification of maximum likelihood function has been constructed as the theoretical basis of this study, make use of Holder inequality and Jensen's inequality to estimate parameters of GARCH(1,1) model with residuals having regularly varying distributions with index 0α>. Finally, the consistency and asymptotic normality of the estimates constructed are further proved.

Mandarin Chinese is known as a tonal language with four lexical tones. Tone recognition plays an important role in automatic Chinese speech recognition in that the same syllable with different tones gives quite distinct meanings. The different tone can be characterized by its pitch contour, but the pitch contours are hardly ideal smooth curves. It is because the pitch points calculated by pitch detector normally have some error points. These error pitch points can cause the erroneous classification of Mandarin four-tone recognition. It is necessary to smooth the pitch contour before tone recognition. The classic smooth algorithms can not deal with error fundamental frequencies successively. A new smoothing method proposed in this paper can deal with the error pitch point appropriately. It first checks whether the current point is a correct or error point, then the error type, and finally modifies the error point according to the error type. For different error type, the corresponding smoothing method is also different. To confirm this smoothing method, four “one vs. all” Support Vector Machine classifier are built for Mandarin Tone Recognition. The test results indicate that error rate of Mandarin Chinese four tone recognition can be reduced under the smoothing method.

According to the image reconstruction accuracy influenced by the “soft field” nature and the limited projection data in electrical capacitance tomography, based on the working principle of the electrical capacitance tomography system, a Novel image reconstruction algorithm based on compressed sensing is proposed in the paper. The method based on ART (algebra reconstruction technique) organically combines the gradient sparse of image and ART, and reduces the norm of image gradient with full-variational method, and improves the accuracy and speed of image reconstruction. Experimental results and simulation data indicate that the imaging accuracy is markedly improved, and the image is closed to the prototype. This new algorithm presents a feasible and effective way to research on image reconstruction algorithm for Electrical Capacitance Tomography System.

Medical image segmentation has important significance for thickness estimation of the articular cartilage and joint disease diagnosis. In this study, a novel automatic segmentation method based on Hough transform and edge detection was proposed to divide femoral cartilage in human hip joint from MR images. MR image was interpolated, smoothed and enhanced in preprocessing to improve the image quality. Hough transform was employed to find out the center position of the femoral head and the anatomical constrain of the hip joint was considered to estimate the area of interest (AOI). Furthermore, the rough segmentation range was extracted. To figure out the border of the cartilage, the adaptive thresholding Canny detector was exerted. The detected edges were then labeled and filtered in a custom one by one manner to remove the noise edges and acquire the exact inner and outer edges of the femoral cartilage, according to the properties of the pixel on femoral cartilage edge. Image data between the two edges were finally extracted to achieve the femoral cartilage segmentation. Experiment on 120 MR image slices proved that the method can automatically segment the femoral cartilage fast and accurately.

As a kind of biometric identification technology Palmprint identification has many characteristics such as large information quantity, high discriminatory power and strong anti-interference ability and so on. The palmprint acquisition and recognition system can be divided into four modules: palmprint image acquisition, preprocessing, feature extraction and feature matching. For completing the hardware design of Nios II processor, image acquisition and recognition module we adopt SOPC Builder, DSP Builder, IDE development tools and Verilog, C programming language. In this system we integrated the verifying algorithm, system design models and variety function interface in a FPGA chip. This method has many obvious advantages such as updatable algorithm, reconfigurable hardware and software and high operational reliability. Experimental results showed that the system meets the real-time authentication and reliable requirement of identification besides it has the reusable ability of system platform.

In the unconstrained environment for video tracking is essential for many applications, such as video surveillance, man-machine interaction. In fact, moving object in the sequences generally has the context information of others or the different moments of its own state. Our research focus on the complex scenes, tracking multiple articulated targets, obtaining the features of the target, getting the precise target segmentation and improving the accuracy and reliability of tracking. We propose using top-down segmentation to feedback object detection, also contains the shape information. And the local appearance information is embedded into the framework of the level set. Then we propose a method to solve the interference of similar appearance target and multi-target tracking, by using context information to create two auxiliary items: Misleading items and support items. Both of them are using continuous random ferns. We experimentally evaluate our proposed approach on challenging sequences and video in real-world demonstrate its good performance in practice.

A simple method of line drawing animation creation method from a human facial video is presented in this paper. Without much manual intervention, the work could be used in personal cartoon video processing and communication. In this domain, a number of works has been proposed, while most of them need interactive processing or specific data such as 3D model. The method presented in this paper could automatically transfer expressions without much user participation. Given an input video and line drawing model with some basic emotions, active appearance models are used to get the current facial feature points of the video, and then the system can achieve the emotion parameters by solving the emotion function of AAM and transfer them to the same basic emotions model of line drawings. The basic emotions of line drawings are formed by background image and facial feature lines which are drawn with a set of Bézier curves with width. Finally, facial line drawing animation is obtained and shown in the experiment.

This paper proposes a novel four-dimensional autonomous system which has complex chaotic dynamics behaviors and gives analysis of novel system. More importantly, the novel system can generate three-layer chaotic attractor, four-layer chaotic attractor, five-layer chaotic attractor, multilayer chaotic attractor by choosing different parameters and initial condition. We analyze the new system by means of phase portraits, Lyapunov exponent spectrum, fractional dimension and bifurcation diagram of the system. The four-dimensional autonomous system is totally different from the well-known systems in previous work. Linear feedback control methods are used to suppress chaos to unstable equilibrium. Numerical simulations are presented to show these results.

The importance of biometric identification technology in the field of information security is increasingly prominent, in various of recognition technology, hand vein recognition technology attracts more and more researchers’ attentions because of its high security and high recognition rate; The traditional template matching method based on vein skeletal morphology inevitably brings about problems such as long training time and too much space occupation of sample storage; the passage applies feature extraction method based on the subspace to the vein recognition on the basis of analysis of the principal component analysis method, which is called (2D)2FPCA algorithm combining the traditional 2DPCA and 2DFLD technology; the new algorithm has many advantages including reduction of the preprocessing algorithm steps and small space occupation of characteristics vectors with high processing speed; Finally, simulation experiments with the new algorithm are carried out in 500 vein image database, which proves that the method not only has better recognition accuracy but also improves the recognition rate while reducing the storage space.

As Global Positioning System(GPS) could provide some navigation information for users such as time, latitude, longitude, and altitude, etc, it has been widely used in various fields like military, communication, measurement and so on. Therefore, it is of great application importance to study the processing method and display way of GPS navigation information. In this paper, on the basis of the positioning principle of GPS, the GPS navigation information processing method is proposed which has combined the PC and the receiver serial communication. The focus of this program is to study the software to process and display the GPS navigation information using the visual programming of VC++. The GPS receiver serial port is used to send real-time navigation information to the PC in this program, and then the time, latitude, longitude, altitude and satellite status and some other information can be displayed in the window frame created by MFC after the extraction of MsComm controls serial port programming. Meanwhile, the received time information can be used to make the simple calendar; the electronic map is loaded with the electronic map control MapX, and then electronic map location function can be achieved using the location information. This design can process and display various GPS navigation information conveniently after the actual test, which can be used in a variety of occasions.

The separation for coal and gangue is an important process in mine production. In order to automatically select the gangue, this study obtained the difference and distribution regularity of the grayscale distribution by analyzing a large number of image data. By improving Bayesian Decision theory, identifiably character Bayesian Discriminant algorithm was proposed to get grayscale division threshold of coal and gangue. Aimed at the problem that impurities gangue and vitrinite in coal affect the accuracy of recognition result, mean smoothing filter algorithm was used to pre-process image and related neighborhood pixels recognition algorithm was proposed for recognizing the coal and gangue. The recognition system was tested on-line with a large number of random selected materials for many times, the average correct recognition was 96.8%. The test results indicated that the algorithm is stable and robust and the recognition system has a great potential in automatic selecting of gangue.

This paper presents two methods for buildings extraction in Very High Resolution (VHR) remotely sensed multispectral (4 band) images based on supervised and unsupervised segmentation using different image properties. The proposed approach for unsupervised or automatic building detection involves four stages, primarily, filtering to smoothen and enhance the objects present and sharpen the details. Secondly, a binary mask creation over which edge detection is applied. Edge linking is done to preserve information about the object. Lastly region properties like area, perimeter, etc are applied on the prepared mask and buildings are detected. The semi-automated or supervised method uses advanced color based segmentation algorithm to extract the buildings tops. This technique creates a number of masks based on segmentation and uses region properties based on color. Experiments are made on VHR images captured from satellites of commercial companies like Digital Globe and Geo Eye. Results of both methods are compared with respect to various accuracy measures at the end. The results illustrate that supervised algorithm using color property produces more accurate object delineation.

Fetal Electrocardiogram (FECG) can provide important clinical information for the assessment of fetal well-being. However, the extraction of FECG from the maternal abdominal wall remains an open problem due to various kinds of interferences and noises, where the maternal Electrocardiogram (MECG) is the dominant source of interference. In this study, a novel FECG extraction framework by combining single-channel singular value decomposition (S-SVD) and cyclostationarity-based blind source separation (C-BSS) is proposed. First, the MECG signal as the principal quasi periodic component is extracted with S-SVD algorithm from a single channel recording. Then, the FECG signal is preliminarily acquired by subtracting the MECG from the corresponding abdominal MECG (AMECG) recording. Finally, the FECG is further extracted by C-BSS, where a new cost function is constructed with high order cumulant and second order cyclic frequency. Results show the proposed method improves the accuracy of extracted FECG in comparison with the traditional BSS algorithm like independent component analysis (ICA).