Search

With the increase in the number of online shopping users, customer loyalty is directly related to product sales. This research mainly explores the statistical modeling and simulation of online shopping customer loyalty based on machine learning and big data analysis. This research mainly uses machine learning clustering algorithm to simulate customer loyalty. Call the k-means interactive mining algorithm based on the Hash structure to perform data mining on the multidimensional hierarchical tree of corporate credit risk, continuously adjust the support thresholds for different levels of data mining according to specific requirements and select effective association rules until satisfactory results are obtained. After conducting credit risk assessment and early warning modeling for the enterprise, the initial preselected model is obtained. The information to be collected is first obtained by the web crawler from the target website to the temporary web page database, where it will go through a series of preprocessing steps such as completion, deduplication, analysis, and extraction to ensure that the crawled web page is correctly analyzed, to avoid incorrect data due to network errors during the crawling process. The correctly parsed data will be stored for the next step of data cleaning or data analysis. For writing a Java program to parse HTML documents, first set the subject keyword and URL and parse the HTML from the obtained file or string by analyzing the structure of the website. Secondly, use the CSS selector to find the web page list information, retrieve the data, and store it in Elements. In the overall fit test of the model, the root mean square error approximation (RMSEA) value is 0.053, between 0.05 and 0.08. The results show that the model designed in this study achieves a relatively good fitting effect and strengthens customers’ perception of shopping websites, and relationship trust plays a greater role in maintaining customer loyalty.

This paper presents an improved fingerprint recognition system for overlapping fingerprint of two fingers. The developed system can automatically and precisely separate an overlapping fingerprint into two areas, an overlapping area and a non-overlapping area, by analyzing their orientation fields and complicities. A method being able to remove a misjudging core of the fingerprint from the overlapping area is also proposed in this paper. Moreover, a recursive correction algorithm and a restrictions relaxation labeling algorithm are applied to separate two fingerprints from the determined overlapping fingerprint area, and the Gabor filter is applied to enhance figure quality of the two separated fingerprints. After that, the VeriFinger 6.2 SDK is used to identify the fingerprints. In the experiments, our system uses the fingerprints from the two public databases. Receiver Operating Characteristic (ROC) and Cumulative Match Characteristic (CMC), are used to examine the proposed fingerprint recognition system. The average recognition rate of our system is 85.7%. The experiments results show the effectiveness of the proposed system.

In recent years, China's e-commerce industry has developed at a high speed, and the scale of various industries has continued to expand. Service-oriented enterprises such as e-commerce transactions and information technology came into being. This paper analyzes the shortcomings and challenges of traditional online shopping behavior prediction methods, and proposes an online shopping behavior analysis and prediction system. The paper chooses linear model logistic regression and decision tree based XGBoost model. After optimizing the model, it is found that the nonlinear model can make better use of these features and get better prediction results. In this paper, we first combine the single model, and then use the model fusion algorithm to fuse the prediction results of the single model. The purpose is to avoid the accuracy of the linear model easy to fit and the decision tree model over-fitting. The results show that the model constructed by the article has further improvement than the single model. Finally, through two sets of contrast experiments, it is proved that the algorithm selected in this paper can effectively filter the features, which simplifies the complexity of the model to a certain extent and improves the classification accuracy of machine learning. The XGBoost hybrid model based on p/n samples is simpler than a single model. Machine learning models are not easily over-fitting and therefore more robust.

This study proposes an obstacle detection method that uses depth information to allow the visually impaired to avoid obstacles when they move in an unfamiliar environment. The system is composed of three parts: scene detection, obstacle detection and a vocal announcement. This study proposes a new method to remove the ground plane that overcomes the over-segmentation problem. This system addresses the over-segmentation problem by removing the edge and the initial seed position problem for the region growth method using the Connected Component Method (CCM). This system can detect static and dynamic obstacles. The system is simple, robust and efficient. The experimental results show that the proposed system is both robust and convenient.

Researchers put efforts to discover more efficient ways to classification problems for a period of time. Recent years, the support vector machine (SVM) becomes a well-popular intelligence algorithm developed for dealing this kind of problem. In this paper, we used the core idea of multi-objective optimization to transform SVM into a new form. This form of SVM could help to solve the situation: in tradition, SVM is usually a single optimization equation, and parameters for this algorithm can only be determined by user’s experience, such as penalty parameter. Therefore, our algorithm is developed to help user prevent from suffering to use this algorithm in the above condition. We use multi-objective Particle Swarm Optimization algorithm in our research and successfully proved that user do not need to use trial – and – error method to determine penalty parameter C. Finally, we apply it to NIST-4 database to assess our proposed algorithm feasibility, and the experiment results shows our method can have great results as we expect.

An ultra-low power, portable, and easily implemented Holter recorder is necessary for patients or researchers of electrocardiogram (ECG). Such a Holter recorder with off-the-shelf components is realized with mixed signal processor (MSP) in this paper. To decrease the complexity of analog circuits and the interference of 60 Hz noise from power line, we use the MSP to implement a finite impulse response (FIR) filter which is equiripple design. We also integrate the ring buffer for the input samples and the symmetrical characteristic of the FIR filter for efficiently computing convolution. The experimental results show that the ECG output signal with the PQRST feature is easy to be distinguished. This ECG signal is recorded for 24 hours using a SD card. Furthermore, the ECG signal is transmitted with a smartphone via Bluetooth to decrease the burden of the Holter recorder. As a result, this paper uses the Lomb method for the spectral analysis of Heart Rate Variability (HRV) better than Fast Fourier Transform (FFT).

At present, the synthesizing faces of different ages does not emphasize on feature alignment and rectification of twisted images. If these situations do happen, they might cause failure and inaccuracy on synthesizing images. In this paper, we propose a reversible human facial aging/rejuvenating synthesis system which is implemented by Active Shape Model (ASM) integrated with Log-Gabor Wavelet, which can be used to search for the dementia elderly. First, we use AdaBoost and ASM algorithm to extract the feature set of human face, and rectify them by the concept of facial geometric invariance. The invariant concepts are the distance between inner corners of both eyes and the distance between the nose and chin. Then, we find manually one target image which is similar to the test image from the database, and analyze age texture of this human image by Log-Gabor wavelet in order to retrieve decomposition maps. Finally, we can effectively simulate human facial images of people of different ages by controlling the number of decomposition map of images and objectively judge the results via the density of wrinkles.

Conventional 2D face recognition methods often struggle when a subject's head is turned even slightly to the side. In this study, a face recognition system based on 3D head modeling that is able to tolerate facial rotation angles was constructed by leveraging the Open source graphic library (OpenGL) framework. To minimize the extensive angle searching time that often occurs in conventional 3D modeling, Particle Swarm Optimization (PSO) was used to determine the correct facial angle in 3D. This reduced the angle computation time to 6 seconds, which is significantly faster than other methods. Experimental results showed that successful ID recognition can be achieved with a high recognition rate of 90%.

Vision based band gesture recognition provides a more nature and powerful means for human-computer interaction. A fast detection process of hand gesture and an effective feature extraction process are presented. The proposed a hand gesture recognition algorithm comprises four main steps. First use Cam-shift algorithm to track skin color after closing process. Second, in order to extract feature, we use BEA to extract the boundary of the hand. Third, the benefits of Fourier descriptor are invariance to the starting point of the boundary, deformation, and rotation, and therefore transform the starting point of the boundary by Fourier transformation. Finally, outline feature for the nonlinear non-separable type of data was classified by using SVM. Experimental results showed the accuracy is 93.4% in average and demonstrated the feasibility of proposed system.

In this paper, we present a physical rehabilitation assistant system based on skeleton detection with Kinect. The users do not have to install the detectors on the exercise equipment anymore. And then, they can just use the rehabilitation equipment with Kinect using the skeleton detection technique. In this study, we build a normalized three-dimensional Cartesian coordinates location of correct postures under OpenNI system. We find out 15 human skeleton joints with three dimensional coordinates and calculate the feature values, than we use support vector machine (SVM) as classifier to define the accuracy of posture. Finally, the system can judge the correct degree of user’s postures. Also, we can have the rehabilitation purpose.

In this paper, a robust and efficient face recognition system based on luminance distribution by using maximum likelihood estimation is proposed. The distribution of luminance components of the face region is acquired and applied to maximum likelihood test for face matching. The experimental results showed that the proposed method has a high recognition rate and requires less computation time.

This paper proposes a novel image inpainting method based on bandelet transform. This technique is based on a multi-resolution layer to perform image restoration, and mainly utilizes the geometrical flow of the neighboring texture of the damaged regions as the basis of restoration. By performing the warp transform with geometrical flows, it transforms the textural variation into the nearing domain axis utilizing the bandelet decomposition method to decompose the non-relative textures into different bands, and then combines them with the affine search method to perform image restoration. The experimental results show that the proposed method can simplify the complexity of the repair decision method and improve the quality of HVS, and thus, repaired results to contain the image of contour of high change, and in addition, offer a texture image of high-frequency variation. These repair results can lead to state-of-the-art results.

Recently, sparse algorithm for signal enhancement is more and more popular issues. In this paper, we apply it to enhance speech signal. The process of sparse theory is classified into two parts, one is for dictionary training part and the other is signal reconstruction part. We focus environment on both white Gaussian noise and color noise filtering based on sparse. The orthogonal matching pursuit (OMP) algorithm is used to optimize the sparse coefficients X of clean speech dictionary, where clean speech dictionary is trained by K-SVD algorithm. Then, we multiply these two matrixes D' and X' to reconstruct the clean speech signal. Denoising performance of the experiments shows that our proposed method is superior to other state of art methods in four kinds of objective quality measures as SNR, LLR, SNRseg and PESQ.

Currently, the automatic image inpainting methods emphasize the inpainting techniques either globally or locally. They didn't consider the merits of global and local techniques to compensate each other. On the contrary, the artists fixed an image in global view firstly, and then focus on the local features of it, when they repaired it. This paper proposes a progressive processing of image inpainting method based on multi-resolution analysis. In damaged and defective area, we imitate the artistic techniques to approach the effectiveness of image inpainting in human vision. First, we use the multi-resolution characteristics of wavelet transform, from the lowest spatial-frequency layer to the higher one, to analyze the image from global-area to local-area progressively. Then, we utilize the variance of the energy of wavelet coefficients within each image block, to decide the priority of inpainting blocks. Finally, we extract the multi-resolution features of each block. We take account of the correlation among horizontal, vertical and diagonal directions, to determine the inpainting strategy for filling image pixels and approximate a high-quality image inpainting to human vision. In our experiments, the performance of the proposed method is superior to the existing methods.

Among all the fingerprint identification/verification systems, such as minutiae-based or filterbank-based fingerprint matching, the performance relies heavily on the quality of the input fingerprint images. In this paper, we propose an effective algorithm of fingerprint image enhancement, which can much improve the clarity and continuity of ridge structures based on the multiresolution analysis of global texture and local orientation by the wavelet transform. Experimental results show that the enhanced image quality by using the wavelet-based enhancement algorithm is much better than the other existing methods for improving the minutiae detection.

We develop an augmented reality (AR) environment with hidden-marker via touch interface using Kinect device, and then also set up a touch painting game with the AR environment. This environment is similar to that of the touch screen interface which allows user to paint picture on a tabletop with his fingers, and it is designed with depth image information from Kinect device setting up above a tabletop. We incorporate support vector machine (SVM) to classify painted pictures which correspond to the inner data and call out its AR into the tabletop in color images information from Kinect device. Because users can utilize this similar touch interface to control AR, we achieve a marker-less AR and interactive environment.

The embedded hybrid fiber-optic/coaxial CATV network can provide cellular mobile communications with a star ring fiber-optic network to distribute broadcasting cable TV programs and to provide the transmission of a bidirectional wireless signal. There are two lasers at the head end. The down-link laser is split into two parts. One part is the local oscillator for the up-link coherent detection, and the other part is externally modulated by the combined TV and wireless signals. The up-link laser is connected to the ring network with the cascaded phase modulators scheme. For the location neighborhood and frequency reuse, we may apply the multistar ring CATV cellular network with the distance between the head end and the first-level hub of 5 km and the distance between the first-level hub and the second-level hub of 0.5 km. The total hub number of 42 for the multistar ring system can support 34,860 wireless users for the IS-54 system with frequency reuse number of 7 covering an area of radius 5 km. © 1995 John Wiley & Sons, Inc.

This paper proposed a bi-directional people-flow counting system with Kinect. It can also be applied to multi-flow to correspond with the demand for the practical application. Firstly, we set the Kinect above the doorway to capture the situation of pedestrian flow. Then this system detects people in the covering area using the depth image information from Kinect system. And we do the morphological processing like erosion to the object and find the region of interest (ROI) often performed on using a mapping-based detection approach. After these previous steps, this system set a detected line and let people go through it. Therefore, we can get people number of the experimental result. For the multi-flow case, it will cause the occlusion problem, so we could apply the depth information to distinguish the target on occlusion problem. Final, we compare the experimental results with the manual count results and other research. Under normal circumstances, our system provides not only almost 100% for bi-directional counting but also correspond with the demand for real-time.1

This paper proposed a new image inpainting method for restoring a large damaged area in the images called the “geometric image inpainting (GII) method”. The method was developed in response to several common problems encountered in the image inpainting process and provided a high-quality result compared to human vision. In the field of image inpainting, there are three key issues: texture analysis, repairing order and reconstruction of lost image pixels. Though there have been a lot of researches published, few combined all three key issues in their studies. We propose a novel image inpainting method integrating all three key issues at the same time. Experiments showed that the proposed image inpainting method offered better results than other existed methods.

This paper presents a novel image hiding method based on multilevel histogram modification and halftoning technique. We use a multilevel histogram modification to finish the data embedding and extracting. The halftone data is extracted from stego-image, and secret image is achieved from LUT inverse halfton technique. Experimental results show that our method provides high capacity of digital image embedding while keeping the distortions at a low level.

At present, the synthesizing faces of different ages does not emphasize on feature alignment and rectification of twisted images. If these situations do happen, they might cause failure and inaccuracy on synthesizing images. In this paper, we propose a reversible human facial aging/rejuvenating synthesis system which is implemented by Active Shape Model (ASM) integrated with Log-Gabor Wavelet, which can be used to search for the dementia elderly. First, we use AdaBoost and ASM algorithm to extract the feature set of human face, and rectify them by the concept of facial geometric invariance. The invariant concepts are the distance between inner corners of both eyes and the distance between the nose and chin. Then, we find manually one target image which is similar to the test image from the database, and analyze age texture of this human image by Log-Gabor wavelet in order to retrieve decomposition maps. Finally, we can effectively simulate human facial images of people of different ages by controlling the number of decomposition map of images and objectively judge the results via the density of wrinkles.

This study proposes a novel image inpainting technique based on watermarking and halftoning. This technique use LSB method to embed error diffusion halftone image into original image for protecting the image. In image repair process, we use LSB method to extract the halftone information, and the reference image is achieved from LUT inverse halftone. Finally we use the reference image to finish the image inpainting work. Experiment shows the performance of our method is very excellent in image inpainting.

In this paper, an efficient face recognition system based on luminance distribution by using maximum likelihood method is proposed. We observe the distribution of the luminance components of the face region and use maximum likelihood scheme to recognize the face. The experimental results show that the proposed method can achieve higher recognition rate efficiently.

This study proposes a novel image inpainting technique based on bandelet transform. This technique using multi-resolution layer to carry out image restoration, it mainly utilizes the geometrical flow of the neighboring texture of the damaged region as the basis of restoration. Firstly, According to these geometrical flows of image to carry out the warp transform, it makes the textural variation transforms into the nearest domain axis. Then utilizes the bandelet decomposition method to decompose the non-related texture into different bands, and combines the affine searching method to carry out the image restoration. The proposed method can simplify the complexity of the repairing decision method, and improve the performance of image inpainting under HVS. Especially for image with many geometry edges, the performance of image inpainting is very remarkable using the proposed technique.

In this paper we present a instant and real-time mobile vehicle license plate recognition system in an open environment. Using a nonfixed video camera installed in the car, the system tries to capture the image of the car in front and to process instant vehicle license plate detection and recognition. Relying on the instant vehicle body recognition, the system can detect and locate the vehicle license plate without the need of background image. Vehicle body detection system utilizes the color characteristics of the barking lights to carry out detection. It first detects the location of the two barking lights in the captured image. Then set license plate detection region using the probability distribution of the license plate between the two lights, thus quickly locate the license plate. This method can eliminate any environmental interference during the license plate detection. From the results of experiment, it is determined that this system can effectively and quickly capture the vehicle image, detect and recognize the license plate whether it is dark, raining or under complicated environments.

Halftone images are widely used in the printing of books, magazines, newspapers and computer printers. To protect the copyright, in this paper, the multi-purpose watermarking system is proposed for halftone images with the dithering technique and particle swarm optimization (PSO). The alternation of dispersed dot order dither cells generated by the PSO achieving good robustness and image quality produce patterns as embedding watermarks. For the purpose of copyright protection, the watermarks are embedded into the halftone image with visually transparency by the variety of dither arrays, robust to malicious attacks and counterfeits. And for the purpose of content authentication, the areas with malicious attacks could be detected and localized by fragile watermarks. In the system, the geometric invariance method, multi-rings Zernike transform (MRZT)containing co-centric rings sub-sampling and the clustering method, are adopted to estimate the angle and scaling factor accurately and achieve geometric invariance. The experimental results show that the watermarking algorithm provides the properties of robust and fragile simultaneously for halftone images.And the watermarked halftone image contains high image quality by the PSO algorithm.

This paper presents an effective method for speaker identification. Based on the wavelet transform, the input speech signal is decomposed into several frequency bands, and then the linear predictive cepstral coefficients (LPCC) of each band are calculated. Furthermore, the cepstral mean normalization technique is applied to all computed features in order to provide similar parameter statistics in all acoustic environments. We propose a multi-band 2-stage vector quantization (VQ) as the recognition model in which different 2-stage VQ classifiers are applied independently to each band and the errors of all 2-stage VQ classifiers are combined to yield total error and a global recognition decision. The experimental results show that the proposed method gives better performance than other recognition models proposed previously in both clean and noisy environments.

This paper presents a fast license plate localization algorithm base on expanded wavelet transform. The system consists of three major stages. First, a wavelet transform based method is used for extracting important contrast features as guides to search for desired license plates. Then, generating candidate regions in LH sub-image. In last stage we locate the license plate accurately in HL sub-image. The 1229 tested images taken from various scenes were employed, including diverse angles, different lightening conditions and the dynamic conditions. The experiments shows that proposed method can quickly and correctly detect the region of license plate and the license plate detecting rate of success is 97.31%. The average processing time of one color image is about 0.016 s.

Currently, noise interference and data loss are two major problems that affect the processing results of image data transmission and storage. In order to restore damaged image data effectively, we propose a novel image inpainting technique based on wavelet transformation. The primary feature of our proposed technique is to separate the given image into two principal components which encompass image texture and color respectively. Then, according to the distinctive qualities of the given image, various image inpainting methods are adopted to perform image repair. By taking advantage of the separation of an image into its individual frequency components, we use the multi-resolution characteristics of wavelet transform, from the lowest spatial-frequency layer to the higher one, to analyze the image from global-area to local-area progressively. In order to substantiate the effectiveness of our proposed image inpainting method, we employed various images subject to high noise interference and/or extensive data loss or distortion. The experimental results were perfect, even if the distortion portions of the repaired images were higher than 90%.

This paper presents a neuro-fuzzy system to speech classification. We propose a multi-resolution feature extraction technique to deal with adaptive frame size. We utilize fuzzy adaptive resonance theory (FART) to cluster each frame. FART was an extension to ART, performs clustering of its inputs via unsupervised learning. ART describes a family of self-organizing neural networks, capable of clustering arbitrary sequences of input patterns into stable recognition codes. In our experiments, the TIMIT database is used and extracts features of each phoneme. The performance of speech classification is 88.66%, demonstrate the effectiveness of the proposed system is encouraging.

Summary form only given. Correct minutiae extraction is very important in an automatic fingerprint identification system. However, the presence of noise in poor-quality images can cause many extraction faults, such as the dropping of true minutiae and inclusion of false minutiae. Most fingerprint identification systems are based on precise mathematical models, but they cannot handle such faults properly. As human beings are good at recognizing fingerprint patterns, a human-like method is applied. The paper presents an adaptive fuzzy logic and neural network method which has variable fault tolerance. Our experimental results show that this fingerprint identification method is robust, reliable and rapid.

This paper presents a wavelet transform based method for extracting license plates from cluttered images. The proposed system consists of three major stages. First, a wavelet transform based method is used for extracting important contrast features as guides to search for desired license plates. Then, finding a reference line in HL subimage plays an important role to locate the desired license plate region roughly. According to the reference line we can decrease the searching region of license plate and speed up the execution time. The last stage of the method is to locate the license plate accurately by license plate adjustment. More importantly, the proposed detection method can locate multiple plates with different orientations in one image. Since the feature extracted is robust to complex backgrounds, the proposed method works well in extracting differently illuminated and oriented license plates. The average accuracy of detection is 92.4%.

This paper presents an effective method for improving the performance of a speaker identification system. Based on the multiresolution property of the wavelet transform, the input speech signal is decomposed into various frequency bands in order not to spread noise distortions over the entire feature space. The linear predictive cepstral coefficients (LPCCs) of each band are calculated. Furthermore, the cepstral mean normalization technique is applied to all computed features. We use feature recombination and likelihood recombination methods to evaluate the task of the text-independent speaker identification. The feature recombination scheme combines the cepstral coefficients of each band to form a single feature vector used to train the Gaussian mixture model (GMM). The likelihood recombination scheme combines the likelihood scores of independent GMM for each band. Experimental results show that both proposed methods outperform the GMM model using full-band LPCCs and mel-frequency cepstral coefficients (MFCCs) in both clean and noisy environments.

In this paper, we present a fast and precise method for fingerprint classification. The proposed method directly extracts the directional information from the thinned image of the fingerprint. We use an octagon mask to search the center point of the region of interest and consider both the direction information and the singular points in the region of interest to classify the fingerprints. In the system, not only is the amount of computation reduced but also can the extracted information be used for identification on AFIS. The system has been tested on the NIST special fingerprint database 4. For the 4000 images in this database, 2000 images are randomly chosen by computer and classified. The classification accuracy reaches 93.1% with no rejection for 4-class classification problem.

In this paper, we propose an effective algorithm of gray-level watermarks based on the characteristics of image. The DCT coefficients of the watermark image are embedded into the position of DCT domain of host image by considering the similarity of DCT blocks energy of both images. The algorithm recovers the watermark without any reference to the original image and we propose a post-processing procedure, which can correct the errors by the survived information, to obtain higher quality watermark. The experimental results show that the proposed algorithm adapted to the original image produces the most imperceptible and the most robust watermarked image under print-and-scanner and various attacks.

With fast development in e-learning, assessment plays an important role between teaching and learning. A good e-learning system has not only good teaching strategy and better learning resources but also a proper assessment model. In this paper, we propose an assessment meta-data and assessment analysis feedback for recent e-learning environments. There is proper feedback for teachers, students, and learning management systems. The feedback could provide proper teaching, learning resource delivery and learning progress suggestions. With this approach, assessment prompts the learning effort in e-learning.

In order to reuse and integrate all resources, many distance learning standards have appeared. The most famous is the Shareable Content Object Reference Model (SCORM). SCORM is composed of a run-time environment (RTE), a content agreement model and LMS. SOAP allows objects or code of any kind, on any platform, in any language to communicate. .Net Web Service combines XML and some programming IDE to develop a server, which uses SOAP to implement the launch mechanism and AICC CMI data model. Java script is used to implement the client side.

Most of the speech segmentation works are based on the thresholds of parameters to segment the speech data into phonemic units or syllabic units. In this paper, we formulate the threshold decision as a clustering problem. Feature parameters extracted from the analysis frame are clustered into three types: silence, consonants, and vowels. Distributed fuzzy rules which have been used in clustering the numerical data are used for this task. The distributed fuzzy rules, which do not need many training data, have good performance in clustering problems and are beneficial for clustering the features of speech data. Such a method, however, has many fuzzy if-then rules. So, we propose a genetic-algorithm-based method for selecting a small number of significant fuzzy if-then rules to construct a compact fuzzy classification system with high classification power. Effectiveness of this approach has been substantiated by classification experiments for continuous radio news speech samples uttered by two females and two males.

A new approach to data clustering which is capable of detecting clusters of different shapes is proposed. In classical clustering approaches, it is a great challenge to separate clusters if the cluster prototypes are difficult to represent by a mathematical formula. In this paper, we propose an improved learning vector quantization (LVQ) algorithm using the concept of symmetry. Through several computer simulations, the results show that the proposed method with random initialization is effective in detecting linear, spherical and ellipsoidal clusters. Besides, this method can also solve the crossed question.

Proposes a simple and efficient embedded image compression algorithm. For embedded image compression, the embedded zerotree wavelet (EZW) method was proposed by J.M. Shapiro (1993), but the EZW method needs a high bit rate to record and to arrange the significant locations in 2D image compression. To obtain a better bit rate and image quality, we propose a new method that is called the embedded important-oriented-tree wavelet (EIW). The EIW bit stream is obtained from the bit stream of the discrete wavelet transform (DWT) by using the correlation of the relative location between sub-band coefficients. By replacing the significant part of the EZW bit stream with the bit stream of the EIW, we don't need to sort, and we can save a lot of bit rate. Simulation results showed that the proposed method obtains a better (lower) bit rate than the EZW method and also produces a fully embedded coding for 2D images.

We propose an idea for enhancing the robustness of extracted watermarks. A watermark can be treated as a transmitted signal, while the destruction of attackers is regarded as noisy distortion in a channel. We suggest an idea using an error correction code (ECC) to detect and correct the error part of the extracted watermarks. In our experiments, the results show that the robustness of a watermark with ECC is much better than the traditional one without ECC.

In a fingerprint recognition system, preprocessing is difficult and wasted time. We propose a method to build a direction matrix which can describe the fingerprint texture to improve the problems of obscure and broken images. Then, utilizing the direction matrix, we propose a feature extraction method to find out counts and locations from the fingerprint singular points, such as core point and delta point. We use NIST(4) as a fingerprint database and the results seem encouraging.

An effective and robust speech feature extraction method is presented. Based on the time-frequency multiresolution property of the wavelet transform, the input speech signal is decomposed into various frequency channels. For capturing the characteristics of an individual speaker, the linear predictive cepstral coefficients of the approximation channel and entropy value of the detail channel for each decomposition process are calculated. In addition, an adaptive thresholding technique for each lower resolution is also applied to remove the influence of noise interference. Experimental results show that using this mechanism not only effectively reduces the influence of noise interference but also improves the recognition performance. Finally, the proposed method is evaluated on the MAT telephone speech database for text-independent speaker identification using the group vector quantisation identifier. Some popular existing methods are also evaluated for comparison, and the results show that the proposed feature extraction algorithm is more effective and robust than the other existing methods. In addition, the performance of the proposed method is very satisfactory even in a low SNR environment corrupted by Gaussian white noise.