Earticle

Objective: The purpose is to acquire EEG signals through an external device, the EEG analysis, calculation of EEG matching values, through calculation and analysis of mobile platform, to provide users with the corresponding help. Method: Three Theory of Feature Extraction are applied to the Brain Matching, that is Short FuLiye Transform Method, Second-Order Blind Identification Method, Phase Synchronization. Conclusion: By moving the platform for EEG, calculate and output, obtain evoked EEG evoked by system design, and the design of EEG feature calculation method, calculated the different match EEG matching values, depending on the application, provide assistance to user, and through the calculation of EEG signal to the user, especially the application of help, friends, compared with other tools more in line with the essential characteristics of the user himself, more authenticity.

Optical communication are difficult to design and analyze therefore it is useful to find the effects of various parameters and characteristics of the Photodetectors used in optical systems. So in this paper effect of various parameters on the performance of proposed system with both types of Photodetectors is analyzed. Through the simulative results taken from “Optisystem 11.0” it is concluded that APD has an edge over PIN at Lower data rates, at lower transmit power levels, at higher attenuation and can support large transmission distance but PIN is preferred at higher thermal noise and higher data rates in comparison to APD.

The proposed RAITIIFNNC system is comprised of a interval type II fuzzy neural network identifier and a robust controller. The identifier is utilized for online estimation of the compound uncertainties. The robust controller is used to attenuate the effects of the approximation error so that the perfect tracking and synchronization of chaotic systems are achieved. All the parameter learning algorithms are derived based on Lyapunov stability theorem to ensure network convergence as well as stable synchronization performance. From the simulation example, to synchronize two Lorenz chaotic systems, it has been shown that the effectiveness of the proposed method has been verified.

This paper introduces a novel solution to generate a super-resolution image from a set of low-resolution input based on patch information. Recent research has shown that super-resolved data can be reconstructed from an extremely small set of measurements compared to that currently required. This paper incorporates the compressive sensing framework to the reconstruction model. Moreover, in order to remove outliers introduced by image parallax, the supervised patch-adaptive matching method which uses photometrical similarity and geometrical distance to determine the matching patch is proposed to reconstruct the high resolution image. The performance of the proposed algorithm on both synthetic and real images is evaluated with several grayscale and color image sequences and found successful when compared to other algorithms.

Mental representation of objects deals with understanding the imagery information of concepts and provides clues for design and implementation of computational object recognition strategies. In this study, I postulate that studying of children’s drawings can provide useful information about mental representation of objects in preliminary stage of developmental learning. To this aim, I have designed an experiment in which I asked children of ages 3-4 to draw objects from known categories. Children were also allowed to have a brief view to a prototypical picture of objects for a short period of time. Based on my observation derived from children's drawings, I classified children’s strategies for object representation into three main categories: part-concept, familiar-concept and shape-concept. Results from this experiment suggest that there is a strong link between early object representation and theories of object recognition.

Detection of starch content in potato is studied applying hyperspectral imaging technique in the paper. The original and preprocessing spectra were processed with partial least square(PLS) method to build prediction model of starch content. The original spectra between 400 and 1000nm was preprocessed with smoothing, second derivation, and multiplicative scatter correction (MSC). Prediction model was built with preprocessing spectra by applying principal component analysis (PCA). Known from the result, the model based on the preprocessing spectra preprocessed with smoothing and PCA is the best of all prediction models built in research. The determination coefficient (R2)of calibration set and prediction set was 0.8234 and 0.9031 respectively. The root mean square error of calibration set (ＲMSEC) and root mean square error of validation set(ＲMSEV) was 0.5633 and 0.5025，respectively．It indicated that this method could be applied in detection of starch content in potato. The study could offer theoretical and practical reference for further study in the future.

Automatic image annotation (AIA) is expected to be a promising way to improve the performance of content-based image retrieval (CBIR). However, current image annotation results are always incomplete and noisy, and far from practical usage. In this paper, we incorporate semantic annotations into CBIR via query expansion scheme to improve retrieval accuracy. In the proposed method, semantic annotations of test images are obtained using a visual nearest-neighbor-based annotation model. And both visual features and annotation keywords are used to represent images. The similarity between two images is determined by their visual similarity and semantic similarity. The method is evaluated on the well-known Pascal VOC 2007 dataset using standard performance evaluation metric. The experimental results indicate that the performance of CBIR can be improved by incorporating semantic annotation via query expansion.

In order to solve the issues like the feature detection algorithm surface design too sensitive to the presence of noise, this thesis proposes the grid processing algorithm. Firstly, the constrained smoothing model is used to smooth the grid, in which the bound terms are the sparse constraint terms of the error term of l2 norm and l1 norm. In the process of smoothing, the points of the smooth location move less, while the there is a greater movement of feature points. And then through the anglicizing the movement distances the initial feature points are extracted; finally, after the generating of the initial feature points, the feature points are more complete. Experimental results show that: the proposed algorithm can handle the noise of too sensitive.

Information is critical in light of the fact that it assists us take a decision. Yet, it needs security. With these worries, picture is the most ideal method for representation of data to to read, write, and and comprehend the data. Face recognition is secure since we can't change our faces, not at all like secret word signature, credit card and debit card that may be abused by others. Appearance, brightening and postures change are the significant testing issues in face acknowledgment. The unwavering quality of face recognition frameworks relies on upon limit of database of facial pictures and testing methodology to assess the face acknowledgment framework. Our examination is concerned with the testing method. This exploration proposed another algorithm of support vector machine. In Experiments we have discovered some tasteful actualities and results. It gives the most noteworthy exactness 97.9 %. This is superior to anything moderately offered results. In the most recent decade, the face recognition framework has advanced with more noteworthy than 90% recognition rate.

To solve the problem of tracking the trajectory of a moving object and learning a deep compact image representation in the complex environment, a novel robust incremental deep learning tracker is presented under the particle filter framework. The incremental deep classification neural network was composed of stacked denoising autoencoder, incremental feature learning and support vector machine to achieve the feature-extracting and classification of particle set. Deep learning is successfully taken to express the image representations obtained effectively. Unsupervised feature learning is used to learn generic image features and transfer learning transforms knowledge from offline training to the online tracking process. The incremental feature learning was consisted of adding features and merging features to online learn compact feature set. Linear support vector machine increases the discretion for target with similar appearance and is further tuned to adapt to appearance changes of the moving object. Compared with the state-of-the-art trackers in the complex environment, the results of experiments on variant challenging image sequences show that incremental deep learning tracker solves the problem of existent trackers more efficiently, it has better robust and more accurate, especially for occlusions, background clutter, illumination changes and appearance changes.

Image mining is one of the data mining research areas and it can be defined as getting hidden information from the image databases. It is used to identify unknown patterns, inherent and valuable information from images. Image mining helps to make relationships between various categories of images which are found in large image databases. These images can reveal useful information to the users. Image mining is distinct from low-level computer vision and image processing techniques. It uses methods from computer vision, image retrieval, image processing, data mining, database, machine learning, and artificial intelligence. Although all these subjects study the same object image, the vital difference between image data mining and the other subjects is, image data mining focuses on large scale set of images while image processing and pattern recognition analysis are based on only single image. Face detection is the problem of determining whether a sub-window of an image contains a face. It has received much attention and has been an extensive research topic in recent years. In this research work, facial images are classified based on its shape feature using optimization algorithms. A new algorithm, i.e. classification based similarity finding is proposed for classifying the facial images as round or oval shape. The performance of the proposed classification based similarity algorithm is compared with the particle swam optimization and genetic algorithms. The results of the existing and proposed algorithms are analyzed based on accuracy and execution time factors. From this we observed that the proposed classification based similarity finding algorithm has produced good results.

Free Space Optical Communication has been developed to cover the longer distances with the growing needs of high speed. It is a data transmission technology from one point to another point by using laser beam carrying information in free space. Multiple transmitters/receivers (TX/RX) are used to improve the performance and quality of Free Space Optics (FSO) communication system under different weather conditions. This paper analyzes the performance of multiple transceivers system by the received power level (PR), Q factor and bit error rate (BER) under clear, fog and haze conditions. From the measured data is possible to calculate visibility which is the main factor for quality and estimation for the availability and reliability of FSO link. The Free Space Optical link was modelled and simulated using a commercial optical system simulator named OptiSystem by Optiwave.

Color is the critical factor in seismic data interpretation and geological targets visualization. And recently, ideas of color blending have brought the enlightenment in attribute combinations for reservoir characterization in petroleum engineering. In this paper, we present this approach of color blending in different color modes and its application in subsurface channel detection by using seismic attributes data. The color models include RGB model, CMY model and HSV model. We firstly calculate sensitive attributes from three dimensional seismic data, including envelop, coherence and spectral decomposition, etc. Then three types of normalized seismic attributes are set as input into the primary color channel of the color models respectively, and then mixed together to create one color blended volume in three dimensional visualization environment. The blended volume has plenty of geological information coming from the three input attributes, resulting in better resolution for channels than the single attribute. Applications in one survey of DQ oilfield show that channels are vividly imaged with special lighted color on the blended volume slices. The spatial distribution characteristics of channels, including the shapes and branches, are clearly depicted. And for the three blending methods, the RGB model is mostly preferred although the CMY model has almost similar performances in channel detection, while HSV model is slightly inferior in this case.

The registration of point cloud is a key problem for model acquisition and 3D reconstruction, it was Put forward that a kind of automatic registration method from multi-views depth image to complete geometric models. First, According to the invariant characteristics of the relative position of the space points in the condition of rigid body transformation, the effective initial matching points array was construct with curvature invariant features and ZNCC, and the coordinate transformation of the matching feature points was solved based on the Unit quaternion, thus the data coarse registration was completed; Then by using fine matching technology based on adaptive distance function and the improved iterative closest surface, the different perspectives of clouds were optimally matched in 3D space; Finally registration error was calculated according to the matching results, and the registration accuracy and speed were analyzed. The results show that, the method can effectively improve the efficiency of registration in the guarantee of the accuracy of registration.

Atmospheric blur is the distortion of image due to long time exposure, fog, wind speed and due to randomly change in refractive index of air through which light travels. Atmospheric blur also occur through non-uniform geometric deformation of image. In this article, we propose a method for restoring atmospheric degraded image using artificial neural network. In proposed methodology use multilayer feed-forward network which trained by error back propagation algorithm and randomly initialize weights of network. This technique provides better result to restore atmospheric blur image and also in the presence of noise.

Due to the advantages of scale invariant feature transform (SIFT) feature points on the invariant to image scale, brightness, rotation, occlusion, noise and so on, this paper proposes a Particle Tracking Velocimetry (PTV) method, based on SIFT feature points matching for velocity measurement of oil-water two-phase flow in horizontal pipelines. The oil-water two-phase flow with large droplet diameter, oil droplets overlap, ununiform lighting, and the centroid position of oil droplets can’t be obtained only by using traditional PTV methods through morphological processing. However, in this paper, the algorithm can directly achieve the average velocity of the flow field according to the positions of correctly matched SIFT feature points, and there is no need to extract the centroid coordinates of each oil droplet. The experimental results show that the proposed algorithm not only can be used in the average velocity measurement of oil-water two-phase flow in horizontal pipelines, but also can reach 95% in the measuring accuracy when the matching feature points are enough sufficient.

Variational Bayesian (VB) inference is the latest iterative method for prediction of data in machine learning. It provides the solution for intractable integration in Bayesian methodology. In this paper, a simple VB linear regression is applied for prediction of the damaged pixels in an image. Bayesian linear regression model is used for prediction of the pixels. For this neighbor pixels are used as training data to generate the parameters of the prediction function. Now using this prediction function, damaged pixels are predicted and incorporated into the image. Proposed method is linear while image is a non-linear object, generally. Hence, for linearity, a small image window size is used to avoid the nonlinearities in image.

In the paper, we present a new noise spectrum estimation algorithm which is simple and effective for non-stationary background noise environments. The new proposed algorithm continuously updates the estimated noise by weighted noisy speech with a constant smoothing factor, the weighting factor is adjusted by an estimated signal-to-noise ratio (SNR), and the SNR is controlled by the local energy which be obtained by frequency smoothing of the noisy power spectrum in each frame. Objective experimental results and a subjective comparison show that the improved noise estimation algorithm when integrated in speech enhancement is preferred over the competitive noise estimation algorithms.

The limited resolution of image sensors and the complex diversity of nature, cause mixed pixel problems in hyperspectral technology. Such problems are common, and increase the complexity of hyperspectral image processing. Hyperspectral unmixing is crucial for hyperspectral image classification and recognition. In unmixing, the image signatures are represented as a linear combination of the basic materials. Unmixing is the process of decomposing a mixed pixel into constituent materials, and calculating the corresponding fractional abundance. If pure materials (end members) are present in an image, unmixing can be divided into two steps, namely, end member extraction and abundance decomposition. On the other hand, if there is no pure material, researchers have devised and investigated unsupervised and semi-supervised spectral unmixing technology. This article presents an overview of the state-of-the-art methods of hyperspectral unmixing and their extensions.

Kinect is a motion-sensing device which was originally developed for the Xbox 360 gaming console. This recently developed low-cost sensor detects the body position, motion, and voice; it consists of a microphone, a RGB camera, and a depth sensor. Kinect is PC-centric sensor which allows developers to develop real-life applications with human gestures and body motions. This paper presents an approach to interpret the indoor room objects in order to match the objects features in depth images captured from an RGBD video database. The dataset consists of color and depth image pairs gathered in real-time indoor home environment. The objects features are matched in depth image pairs with the feature association method to detect stable features at different time instances.

Vector magnitude calculation is significant in the QR-algorithm by Coordinate Rotation Digital Computer (CORDIC), which is increasingly used in adaptive applications. However, its hardware implementation is extremely difficult. In this paper, a fixed point CORDIC system is constructed to compute the magnitude of a vector by using Model-Based Design. The simulation results show that the proposed method of CORDIC vector magnitude calculator is not only simple in structure and easy to implement, and can improve the speed of operation with good scalability.

Based on Kinect platform, human’s gesture recognition and imitation were realized by humanoid robot NAO in this paper. The hardware system structure of Kinect platform and the principle of human skeleton extraction were mainly introduced firstly. From the image, Kinect camera gets the skeletal point information which is used to compared with preset posture of the skeletal information. Through calculate the angle between the different bones, the computer obtains the human’s posture and sends the instructions to the robot. In the physical experiments, NAO robot receives the pre-programmed corresponding action instructions and imitates the corresponding motions of humans.

This paper shows the study of free vibration analysis of stiffened isotropic plates with orthogonal stiffeners placed eccentrically and concentrically to the plates. In this paper finite element model is developed in ANSYS parametric design language code and discretized using 20 node structural element (SOLID 186) and convergence study of isotropic stiffened plates has been performed and compare the results with related published literature. Effects of various parameters such as boundary conditions, aspect ratios, position of stiffeners eccentrically and concentrically to the stiffened plates has been studied. The vibration analysis of stiffened plate have been studied using Block -Lanczos algorithm. The results of non dimensional frequency of eccentric and concentric isotropic stiffened plate have been compare at different mode shapes, aspect ratio’s, boundary conditions using ANSYS.

A multiple moving object segmentation algorithm based on Background Modeling and Adaptive Clustering (named as BMAC) algorithm is proposed in this paper. For moving object segmentation, the algorithm uses Chebyshev inequality and the kernel density estimation method to do background modeling firstly. Then in order to classify image pixels as background points, foreground points and suspicious points, an adaptive threshold algorithm is proposed accordingly. After using background modeling, adaptive clustering is used for multi-object segmentation. It defines pixel space connectivity rate and designs a perpendicular split method, initial cluster adaptive splitting and merging self-organizing the iterative clustering segmentation algorithm, without pre-set number of clustering, completes multi-object segmentation for the foreground image. The segmentation results are consistent with the human visual judgment, the use of space connectivity information improve the accuracy of clustering segmentation, comparison and analysis the experimental results show that the proposed algorithm is feasible, rapid and effective.

In order to expand the dynamic range of the GMI sensor in longitudinally excitated amorphous wire and improve its precision, waveforms of the GMI sensor are analyzed on the background of weak magnetic field measurement. Then three features extraction methods are studied in detail. According to the advantages and disadvantages of different methods, an improved method which combines the energy features of the wavelet decomposition and the amplitude features is proposed. First, fit the amplitude change ratio curve respectively with Gaussian function and polynomial function, which not only solves the problem of nonlinearity, but also improves the measurement accuracy. Considering the difference of signals’ in-pulse features at different positions, the ‘db5’ wavelet is introduced to decompose the signals. Then the BP neural network trained by the energy features of the wavelet is used to locate the target’s approximate position, as a result, the problem of multi-value is solved. At last, experiments of target detection in weak magnetic field prove that the method proposed is effective.

With the rise in demand for channel capacity, Wavelength Division Multiplexing (WDM) technique is employed in optical communication networks. Gain Flattened Erbium Doped Fiber Amplifier (EDFA) is used in optical WDM system to obtain a uniform output power level. The performance of EDFA depends on various parameters such as fiber length, pump power, pumping wavelength and pumping technique. The objective of this paper is to analyze the performance of EDFA for different pumping techniques namely co-propagating, counter-propagating and bidirectional propagating in a 32 channel C-band optical WDM transmission system at different levels of pump power and different fiber lengths. The system has been analyzed on the basis of received power, BER, gain, noise figure and Q- Factor in the wavelength range of 1528nm to 1562nm at -25dBm transmission power and 0.5nm channel spacing. The results obtained using different pumping techniques are compared to find an appropriate pumping technique which optimizes the performance of the optical WDM transmission system.

A human facial expression is the formation of facial muscle movement. In our previous research, we proposed a method of identifying facial muscle movement which based on motion templates and GentleBoost. But the method was not robust enough to recognize human expression due to insufficient learning stage. So in this paper, we proposed a new method based on motion templates and 4-layer deep learning neural network to identify human's facial expressions. We recognized Action Unit as a kind of features by using motion templates and adaboost firstly, and then the extracted features were used to feed a 4-layer deep learning neural network to recognize the facial expression. The experimental results have proved that the proposed method can solve the problem encountered in our previous research.