Search

This paper presents a novel adaptive pipelined neural finite impulse response (PNFIR) filter for nonlinear signal processing. Unlike traditional pipelined recurrent neural network (PRNN), each module of the PNFIR filter is a simple architecture that includes a standard FIR filter followed by a nonlinear activation function. The complete design of proposed filter includes two subsections: The nonlinear part consists of neural FIR (NFIR) modules which is interconnected in a chained form and simultaneously executed in a parallel fashion; the linear subsection is a tapped-delay-line (TDL) linear combiner. Based on convex combination architecture, the adaptive algorithm derived from the gradient descent approach is utilized to update weights of the nonlinear and linear parts. Moreover, the analysis of stability conditions and computational complexity is also presented. Numerous simulation experimental results on nonlinear dynamic systems identification, speech signal and chaotic time series prediction show that the proposed PNFIR filter has simpler architecture, faster convergence rate, and lower computation complexity than the PRNN and joint process filter using pipelined feedforward second-order Volterra architecture (JPPSOV).

Regularization parameter selection (RPS) is one of the most important tasks in solving inverse problems. The most common approaches seek the optimal regularization parameter (ORP) from a sequence of candidate values. However, these methods are often time-consuming because they need to conduct the estimation process on all candidate values, and they are always restricted to solve certain problem types. In this paper, we propose a novel machine learning-based prediction framework (MLBP) for the RPS problem. The MLBP first generates a large number of synthetic data by varying the inputs with different noise conditions. Then, MLBP extracts some pre-defined features to represent the input data and computes the ORP of each synthetic example by using true models. The pairs of ORP and extracted features construct a training set, which is used to train a regression model to describe the relationship between the ORP and input data. Therefore, for newly practical inverse problems, MLBP can predict their ORPs directly with the pre-trained regression model, avoiding wasting computational resources on improper regularization parameters. The numerical results also show that MLBP requires significantly less computing time and provides more accurate solutions for different tasks than traditional methods. Especially, even though the MLBP trains the regression model on synthetic data, it can also achieve satisfying performance when directly applied to field data.

To reduce the computational burden of the generalized FLANN (GFLANN) filter for nonlinear active noise control (NANC), a hierarchical partial update GFLANN (HPU-GFLANN) filter is presented in this paper. Based on the principle of divide and conquer, the proposed HPU-GFLANN divides the complex GFLANN filter (i.e., long memory length and large cross-terms selection parameter) into simple small-scale GFLANN modules and then interconnected in a pipelined form. Since those modules are simultaneously performed in a parallelism fashion, there is a significant improvement in computational efficiency. Besides, a hierarchical learning strategy is used to avoid the coupling effect between the nonlinear and linear part of the pipelined architecture. Data-dependent hierarchical M-Max filtered-error LMS algorithm is derived to selectively update coefficients of the HPU-GFLANN filter, which can further reduce the computational complexity. Moreover, the convergence analysis of the NANC system indicates that the proposed algorithm is stable. Computer simulation results verify that the proposed adaptive HPU-GFLANN filter is more effective in nonlinear ANC systems than the FLANN and GFLANN filters.

A stochastic analysis of the probabilistic least-mean-square (Prob-LMS) algorithm would be a useful guideline for designing the adaptive filter. However, no analytical expressions for the stochastic analysis of the Prob-LMS algorithm have been reported in the literature. Hence, this paper analyzes the mean-deviation and mean-square-deviation (MSD) behavior of the Prob-LMS algorithm for the general case of an unknown Gauss-Markov channel. Analytical expressions are derived for the transient and steady-state MSD of the Prob-LMS algorithm. Monte Carlo simulations for fixed and time varying channels show excellent agreement between the simulated and theoretical MSD for a wide range of parameters such as SNR, filter length and input signal statistics. Monte Carlo MSD simulation results are presented for the Prob-LMS algorithm which compare favorably to several well-known VSS algorithms.

Non-intrusive methods for eye tracking are important for many applications of vision-based human computer interaction, such as driver fatigue detection, eye gaze replacing the hand operating mouse, eye typing instead of manually depressing keys as a virtual keyboard, eye gaze correction for video conferencing, interactive assistant application for disabled users, etc. However, due to the eye motion be the high nonlinearity, the obstacles of robustness of external interference and accuracy of eye tracking, these tend to significantly limit their scope of application. In this paper, we present a strong tracking finite-difference extended Kalman filter algorithm, and overcome the modeling of nonlinear eye tracking. In filtering calculation, strong tracking factor is introduced to modify prior covariance matrix to improve the accuracy of the filter. The filter uses finite-difference method to calculate partial derivatives of nonlinear functions to eye tracking. The last experimental results show validity of our method for eye tracking under realistic conditions.

Objectives Spontaneous intracerebral hemorrhage remains a major cause of death and disability throughout the world. We tried to establish accurate long‐term outcome prediction models for hypertensive intracerebral hemorrhage (HICH) using CT radiomics and machine learning. Methods In a retrospective study of 270 patients with HICH between June 2013 and June 2018, CT images and patients' 6‐month outcome based on the modified Rankin Scale were collected. Hematomas on CT images were selected as volumes of interests (VOIs), and 1,029 radiomics features of the VOIs were extracted. Based on correlations with patients' outcome, radiomics features underwent dimensionality reduction analyses. Then, the support vector machine (SVM), k‐nearest neighbor (KNN), logistic regression (LR), decision tree (DT), random forest (RF), and XGBoost algorithms were applied with the screened features to establish prognostic prediction models of HICH. Accuracies of all models were compared. Results Eighteen radiomics features were screened as prognosis‐associated radiomics signature of HICH based on the variance threshold, SelectKBest, and least absolute shrinkage and selection operator (LASSO) regression models. Patients were randomly allocated into training (n = 215) and validation (n = 55) sets. Accuracies of all 6 machine learning algorithms in the validation set exceeded 80%. The sensitivity, specificity, and accuracy in the validation set were 93.3%, 92.5%, and 92.7% for the RF model and 92.3%, 88.1%, and 89.1% for the XGBoost model, respectively, which were the best two among all models. Conclusions Taking advantage of radiomics and machine learning, we established accurate prognostic prediction models of HICH. The RF model and XGBoost model returned the best accuracies., We tried to establish accurate long‐term outcome predictions models for hypertensive ICH using CT radiomics and machine learning. Six machine learning algorithms were used at the same time, and the accuracies of different models were compared and selected. Finally, the random forest algorithm model demonstrated a sensitivity of 93.3%, a specificity of 92.5% and the area under the curve was 0.92, which were the best among all 6 models.

We present a novel face recognition method based on direct discriminant Volterra kernels and effective feature classification (DD-VK). One of the crucial steps involves dividing face images into patches and using the DD-VK method to extract the features of sub-image patches. DD-VK implements diagonalization to discard useless information in the null space of the inter-class scatter matrix and preserve important discriminant information in the null space of the intra-class scatter matrix. This method can simultaneously maximize inter-class distances and minimize intra-class distances in the feature space. We also introduce a novel classification scheme associated with the 2D Volterra kernel feature. Our scheme aggregates the classification information obtained from each column of the feature matrix in each image patch and uses a voting strategy to implement parent face image classification. This procedure can reduce the influence of local negative information. Experimental results show that the proposed method demonstrates good performance when dealing with conventional face recognition problems and exhibits strong robustness when dealing with block occlusion images.

In real life, many engineering problems are modeled in the complex domain (CD). This paper proposes two complex-valued proportionate adaptive filtering (CPAF) algorithms in the CD for identifying the complex sparse systems in impulsive noises. The proposed CPAF algorithms are derived by maximizing the reuse of Versoria cost subjected to the weighted squared Euclidean norm of the filter tap-weight vector difference with the proportionate matrices as the weight of the weighted Euclidean norm. To address the tradeoff problem between the fast convergence rate and low steady-state misadjustment of the CPAF algorithms, two combined-step-size (CSS) CPAF algorithms are also proposed in the CD, which are obtained by introducing a time-varying step-size bound (TVSSB) into the constraint condition of the cost function of the CPAF algorithms. The TVSSB is achieved adaptively by using a modified sigmoidal function (MSF) to combine a large step-size and a small one. In order to be robust against impulsive noises, the MSF is adapted with the aid of a stochastic gradient ascent algorithm which is obtained by maximizing the Versoria cost function with respect to the real part of the a priori prediction error. Simulations on complex sparse system identification under impulsive noises in the CD demonstrate that the performance of the proposed CPAF algorithms is superior to that of the complex-valued variants of the affine projection Versoria, affine projection sign (APS), proportionate APS (PAPS), and improved PAPS (IPAPS) algorithms, also validate the effectiveness of the proposed CSS-CPAF algorithms.

The classical proportionate adaptive filtering (PAF) algorithms achieve a fast initial convergence for sparse impulse response. But the small coefficients receive very little gain so that the time needed to reach steady-state misalignment is increased. In addition, the PAF algorithms converge much slower than the original adaptive filtering (OAF) algorithms when the impulse response is dispersive. In order to address these problems, this paper proposes a family of gain-combined PAF (GC-PAF) algorithms. The gain-combined matrix of the proposed GC-PAF algorithms is implemented by using a sigmoidal activation function to adaptively combine the proportionate matrix and identity matrix, which can retain the advantages of both the PAF algorithms in the context of sparse impulse response and the OAF algorithms in the context of dispersive impulse response. Meanwhile, to be also applicable to the family of sign algorithms against impulsive noise, a general framework for the update of the sigmoidal activation function is obtained by using the gradient descent method to minimize the L1-norm of the system output error. Simulations in the contexts of three different sparsity impulse responses have shown that the proposed GC-PAF algorithms perform much better than the OAF, PAF and improved PAF (IPAF) algorithms.

The pipelined adaptive Volterra filters (PAVFs) with a two-layer structure constitute a class of good low-complexity filters. They can efficiently reduce the computational complexity of the conventional adaptive Volterra filter. Their major drawbacks are low convergence rate and high steady-state error caused by the coupling effect between the two layers. In order to remove the coupling effect and improve the performance of PAVFs, we present a novel hierarchical pipelined adaptive Volterra filter (HPAVF)-based alternative update mechanism. The HPAVFs with hierarchical decoupled normalized least mean square (HDNLMS) algorithms are derived to adaptively update weights of its nonlinear and linear subsections. The computational complexity of HPAVF is also analyzed. Simulations of nonlinear system adaptive identification, nonlinear channel equalization, and speech prediction show that the proposed HPAVF with different independent weight vectors in nonlinear subsection has superior performance to conventional Volterra filters, diagonally truncated Volterra filters, and PAVFs in terms of initial convergence, steady-state error, and computational complexity.

In this paper, the robust Bounding Ellipsoidal Adaptive Constrained least-squares (BEACON) algorithm based on new set-membership error bounds is proposed to improve robustness and steady-state misalignment in impulsive noise environments. The expressions for the steady-state excess mean square error (EMSE) of the proposed robust BEACON are obtained. Simulations on system identification and double-talk scenarios show that the robust BEACON algorithm considerably outperforms the traditional BEACON algorithm in the presence of impulsive noise and the theoretical predictions are in good agreement with simulation results.

In this paper, we proposed a novel real-time mosaic block detection method based on intensity order and shape feature to automatically detect mosaic blocks in video. In the proposed method, the video frame is firstly converted to gray image, and it is divided into several layers according to the gray-scale levels to obtain the binary images. Secondly, performed the morphological operations on binary image include dilation and erosion to exclude the influence of single pixel and small area. Finally, extracted the shape features of connected domain of each binary image are used to detect the mosaic block. Experiments show that the proposed method can effectively detect the abnormal image which containing mosaic blocks in video and has the good performance on accuracy.

In order to enhance the classification accuracy of the two-dimensional feature of the image, the idea of a separate classification for each projection direction feature is proposed in this paper. Our method first divides the image into blocks and finds the two-dimensional sub-projection matrix of each sub-block, and then completes the feature extraction by using each column of the projection matrix. At the stage of feature classification, the feature of each projection direction of each sub-block is classified separately, cast a vote towards images classification and the result with the maximum votes wins. We named this method decision level vote classification (DL_VC). The proposed method is equivalent to transforming global features into local features, and the voting method can reduce the influence of local interference. Experiments on face databases demonstrate the efficacy of the proposed method.

With its three vertically stacked photodiodes, the Foveon X3 image sensor enables direct captures of red, green and blue light at each pixel in an image during a single exposure. It greatly reduces computational time for interpolation for missing colors in conventional Bayer sensors and therefore color aliasing. Meanwhile, with the new Super-Fine Detail (SFD) exposure mode, cameras equipped with Foveon X3 technology can capture as many as 51 megapixel-equivalent noiseless images with an extensive dynamic range.

The main drawback of the recursive least p-norm (RLpN) adaptive-filtering algorithm is a poor tracking performance in the presence of abrupt changes in the model. In this paper, a new method to enhance tracking capability of the RLpN (ET-RLpN) algorithm is proposed, which uses the adaptive gain factor in the cross-correlation vector and the input-signal autocorrelation matrix to enhance tracking capability. Simulation results in system identification and echo cancellation applications are presented, which demonstrate that the ET-RLpN achieves improved tracking capability compared to the conventional RLpN and controlled adaptive combination of two RLpN filters (CAC-RLpN).

This paper presents a novel approach based on enhanced local directional patterns (ELDP) to face recognition, which adopts local edge gradient information to represent face images. Specially, each pixel of every facial image sub-block gains eight edge response values by convolving the local 3x3 neighborhood with eight Kirsch masks, respectively. ELDP just utilizes the directions of the most prominent edge response value and the second most prominent one. Then, these two directions are encoded into a double-digit octal number to produce the ELDP codes. The ELDP dominant patterns (ELDP^d) are generated by statistical analysis according to the occurrence rates of the ELDP codes in a mass of facial images. Finally, the face descriptor is represented by using the global concatenated histogram based on ELDP or ELDP^d extracted from the face image which is divided into several sub-regions. The performances of several single face descriptors not integrated schemes are evaluated in face recognition under different challenges via several experiments. The experimental results demonstrate that the proposed method is more robust to non-monotonic illumination changes and slight noise without any filter.

The existing palmprint protection schemes without verification degrading do not strictly meet non-invertible requirement; while non-invertible palmprint protection schemes do not meet verification performance requirement in stolen-token scenario. PalmHash Code and PalmPhasor Code, as two cancelable palmprint coding schemes, are proposed in this paper to balance the conflict between security and verification performance. In order to reduce computational complexity and storage cost, cancelable palmprint coding frameworks are extended from one dimension to two dimensions. Besides, two measures, perpendicular orientation transposition and multi-orientation score level fusion, are employed to further improve the performance of two dimensional (2D) cancelable palmprint codes. PalmHash Code and PalmPhasor Code are compared extensively in this paper. The experimental results and analysis confirm that 2D PalmHash Code and 2D PalmPhasor Code enhanced by the two measures are effective and secure even in stolen-token scenario.

Discriminant locality preserving projection (DLPP) is a linear approach that encodes discriminant information into the objective of locality preserving projection and improves its classification ability. To enhance the nonlinear description ability of DLPP, we can optimize the objective function of DLPP in reproducing kernel Hilbert space to form a kernel-based discriminant locality preserving projection (KDLPP). However, KDLPP suffers the following problems: 1) larger computational burden; 2) no explicit mapping functions in KDLPP, which results in more computational burden when projecting a new sample into the low-dimensional subspace; and 3) KDLPP cannot obtain optimal discriminant vectors, which exceedingly optimize the objective of DLPP. To overcome the weaknesses of KDLPP, in this paper, a direct discriminant locality preserving projection with Hammerstein polynomial expansion (HPDDLPP) is proposed. The proposed HPDDLPP directly implements the objective of DLPP in high-dimensional second-order Hammerstein polynomial space without matrix inverse, which extracts the optimal discriminant vectors for DLPP without larger computational burden. Compared with some other related classical methods, experimental results for face and palmprint recognition problems indicate the effectiveness of the proposed HPDDLPP.

Based on non-linear Volterra kernels mapping and direct discrimination analysis(DD-Volterra), a novel face recognition algorithm is proposed. Firstly, the original image is segmented into specific sub blocks and seeks functional mapping using truncated Volterra kernels. Next, simultaneous diagonalization obtain Volterra kernel optimal projection matrix. This matrix can discard useless information that exist in the null space of the inter-class. Also, it can reserve discriminative information that exist in the null space of the intra-class. Finally, in the test, each block of the test image is classified separately, voting strategy and nearest neighbor classifier algorithm are used for classification. Experiments show that the proposed DD-Volterra method has better performance for it is more effective than Volterrafaces during the extracting facial feature stage.

In this paper, we propose a novel approach for palmprint recognition, which contains two interesting components: directional representation and compressed sensing. Gabor wavelets can be well represented for biometric image for their similar characteristics to human visual system. However, these Gabor-based algorithms are not robust for image recognition under non-uniform illumination and suffer from the heavy computational burden. To improve the recognition performance under the low quality conditions with a fast operation speed, we propose novel palmprint recognition approach using directional representations. Firstly, the directional representation for palmprint appearance is obtained by the anisotropy filter, which is robust to drastic illumination changes and preserves important discriminative information. Then, the principal component analysis (PCA) is used for feature extraction to reduce the dimensions of the palmprint images. At last, based on a sparse representation on PCA feature, the compressed sensing is used to distinguish palms from different hands. Experimental results on the PolyU palmprint database show the proposed algorithm have better performance than that of the Gabor based methods.

In this paper, we present the performance analysis of a self-recovery fragile watermarking scheme using block-neighbor- hood tamper characterization. This method uses a pseudorandom sequence to generate the nonlinear block-mapping and employs an optimized neighborhood characterization method to detect the tampering. Performance of the proposed method and its resistance to malicious attacks are analyzed. We also investigate three optimization strategies that will further improve the quality of tamper localization and recovery. Simulation results demonstrate that the proposed method allows image recovery with an acceptable visual quality (peak signal-to-noise ratio (PSNR) as 25 dB) up to 60% tampering.

This paper presents a novel supervised linear dimensionality reduction approach called maximum margin neighborhood preserving embedding (MMNPE). The central idea is to modify the neighborhood preserving embedding by maximizing the maximum margin distance while preserving the geometric structure of the manifold. Experimental results conducted on the ORL database, the Yale database and the VALID face database indicate the effectiveness of the proposed MMNPE.

To compensate the linear and nonlinear distortions and to track the characteristic of the time-varying channel in digital communication systems, a novel adaptive decision feedback equalizer (DFE) with the combination of finite impulse response (FIR) filter and functional link neural network (CFFLNNDFE) is introduced in this paper. This convex nonlinear combination results in improving the convergence speed while retaining the lower steady-state error at the cost of a small increasing computational burden. To further improve the performance of the nonlinear equalizer, we derive here a novel simplified modified normalized least mean square (SMNLMS) algorithm. Moreover, the convergence properties of the proposed algorithm are analyzed. Finally, computer simulation results which support analysis are provided to evaluate the performance of the proposed equalizer over the functional link neural network (FLNN), radial basis function (RBF) neural network and linear equalizer with decision feedback (LMSDFE) for time-invariant and time-variant nonlinear channel models in digital communication systems.

This paper presents a computationally efficient nonlinear adaptive filter by a pipelined functional link artificial decision feedback recurrent neural network (PFLADFRNN) for the design of a nonlinear channel equalizer. It aims to reduce computational burden and improve nonlinear processing capabilities of the functional link artificial recurrent neural network (FLANN). The proposed equalizer consists of several simple small-scale functional link artificial decision feedback recurrent neural network (FLADFRNN) modules with less computational complexity. Since it is a module nesting architecture comprising a number of modules that are interconnected in a chained form, its performance can be further improved. Moreover, the equalizer with a decision feedback recurrent structure overcomes the unstableness thanks to its nature of infinite impulse response structure. Finally, the performance of the PFLADFRNN modules is evaluated by a modified real-time recurrent learning algorithm via extensive simulations for different linear and nonlinear channel models in digital communication systems. The comparisons of multilayer perceptron, FLANN and reduced decision feedback FLANN equalizers have clearly indicated the convergence rate, bit error rate, steady-state error and computational complexity, respectively, for nonlinear channel equalization.

Due to the limitation of the storage space in the real-world face recognition application systems, only one sample image per person is often stored in the system, which is the so-called single sample problem. Moreover, real-world illumination has impact on recognition performance. This paper presents an illumination robust single sample face recognition approach, which utilizes multi-directional orthogonal gradient phase faces to solve the above limitations. In the proposed approach, an illumination insensitive orthogonal gradient phase face is obtained by using two vertical directional gradient values of the original image. Multi-directional orthogonal gradient phase faces can be used to extend samples for single sample face recognition. Simulated experiments and comparisons on a subset of Yale B database, Yale database, a subset of PIE database and VALID face database show that the proposed approach is not only an outstanding method for single sample face recognition under illumination but also more effective when addressing illumination, expression, decoration, etc.

This paper presents a challenge/response-based biometric image scrambling scheme which attempts to overcome the liveness and retransmission issues of biometric image transmission over insecure communication channel. An intelligent biometric sensor is proposed which has computational power to receive challenges from authentication server and generates response against the challenge with encrypted biometric image. The fractional Fourier transform (FRT) has been used for a biometric image encryption process with its scaling factors and random phase masks as the secret keys. The random phase masks of FRT are chaotically generated by piecewise linear chaotic map (PWLCM) to further improve the encryption security. Experimental and simulation results have demonstrated that the proposed system is secure, robust, and deters the risks of attacks for secure biometric image transmission.

This paper, presents a novel unsupervised dimensionality reduction approach called variance difference embedding (VDE) for facial feature extraction. The proposed VDE method is derived from maximizing the difference between global variance and local variance, so it can draw the close samples closer and simultaneously making the mutually distant samples even more distant from each other. VDE utilizes the maximum variance difference criterion rather than the generalized Rayleigh quotient as a class separability measure, thereby avoiding the singularity problem when addressing the sample size problem. The results of the experiments conducted on ORL database, Yale database and a subset of PIE database indicate the effectiveness of the proposed VDE method on facial feature extraction and classification.

Recently, Xiao et al. proposed an improved key agreement protocol based on chaotic maps, in which only a predetermined long-term key is utilized to ensure its security. This paper demonstrates that none of these schemes can satisfy the contributory nature of key agreement. To fill the gaps, we present a secure key agreement protocol based on chaotic Hash. The proposed scheme utilizes the chaotic Hash function to achieve the contributory nature and enhance its security. Cryptanalysis demonstrates that our chaotic Hash-based scheme can overcome all the current deficiencies.

A novel scheme for generating cancelable palmprint templates is proposed in this paper. Firstly, a chaotic high speed stream cipher is implemented based on coupled nonlinear dynamic filters (CNDF), in which the CNDF are constructed to have flows inverse to each other. Secondly, renewable and privacy preserving palmprint templates are generated using the CNDF chaotic stream cipher with multiple orientation palmprint features obtained from a bank of Gabor filters and encoded in a phase-coding scheme. Compared with the standard palmprint templates, the cancelable templates have greater ability to discriminate palmprints from different hands by increasing the inter-class divergence of different palms more effectively, while maintaining the intra-class distance among palmprints of the same hands. Lastly, the matching stage is performed directly on the cancelable/encryption domain in parallel to accelerate matching and to protect user's privacy. Several fusion rules are investigated for the matching scores of different directional PalmCodes to obtain the final matching score. Compared with Max, Min, Median and Product fusion rules, the Sum rule can greatly accelerate the speed and improve the performance Experimental results on the Hong Kong PolyU Palmprint Database verify that the proposed cancelable templates can achieve very high performance and security levels with a very strong ability to reissue palmprint templates. The proposed method can also be implemented at high speed, which satisfies the needs of real-time applications.

In this paper we propose a two-dimensional multiple maximum scatter difference (2DMMSD) method for face representation and recognition. The new algorithm is based on multiple maximum scatter difference (MMSD) discriminant criterion and image matrices projection technique. The 2DMMSD method does not need to transform image matrix into a vector prior to feature extraction so that it is computationally more efficient and accurate than MMSD in extracting the facial features. Experimental results on ORL and Yale face databases demonstrate the effectiveness and robustness of the proposed method.

A new scheme for driver fatigue detection is presented, which is based on the nonlinear unscented Kalman filter and eye tracking. Assuming a probability distribution than to approximate an arbitrary nonlinear function or transformation, eye nonlinear tracking can be achieved using an unscented transformation (UT), which adopts a set of deterministic sigma points to match the posterior probability density function of the eye movement. Driver fatigue can be detected using the percentage of eye closure (PERCLOS) framework in a realistic driving condition after the eye nonlinear tracking. This system was tested adequately in realistic driving environments with subjects of different genders, with/without glasses, in day/night driving, being commercial/noncommercial drivers, in continuous driving time, and under different road conditions. The last experimental results show that the proposed method not only improves the robustness for nonlinear eye tracking, but also can provide more accurate estimation than the traditional Kalman filter.

Non-intrusive methods for eye tracking are important for many applications of vision-based human computer interaction. However, due to the high nonlinearity of eye motion, how to ensure the robustness of external interference and accuracy of eye tracking pose the primary obstacle to the integration of eye movements into today’s interfaces. In this paper, we present a strong tracking unscented Kalman filter (ST-UKF) algorithm, aiming to overcome the difficulty in nonlinear eye tracking. In the proposed ST-UKF, the Suboptimal fading factor of strong tracking filtering is introduced to improve robustness and accuracy of eye tracking. Compared with the related Kalman filter for eye tracking, the proposed ST-UKF has potential advantages in robustness and tracking accuracy. The last experimental results show the validity of our method for eye tracking under realistic conditions.

A novel nonlinear adaptive filter with pipelined Chebyshev functional link artificial recurrent neural network (PCFLARNN) is presented in this paper, which uses a modification real-time recurrent learning algorithm. The PCFLARNN consists of a number of simple small-scale Chebyshev functional link artificial recurrent neural network (CFLARNN) modules. Compared to the standard recurrent neural network (RNN), those modules of PCFLARNN can simultaneously be performed in a pipelined parallelism fashion, and this would lead to a significant improvement in its total computational efficiency. Furthermore, contrasted with the architecture of a pipelined RNN (PRNN), each module of PCFLARNN is a CFLARNN whose nonlinearity is introduced by enhancing the input pattern with Chebyshev functional expansion, whereas the RNN of each module in PRNN utilizing linear input and first-order recurrent term only fails to utilize the high-order terms of inputs. Therefore, the performance of PCFLARNN can further be improved at the cost of a slightly increased computational complexity. In addition, due to the introduced nonlinear functional expansion of each module in PRNN, the number of input signals can be reduced. Computer simulations have demonstrated that the proposed filter performs better than PRNN and RNN for nonlinear colored signal prediction, nonstationary speech signal prediction, and chaotic time series prediction.

Statistical fragile watermarking is capable of accurately locating tampered pixels and resisting the collage attack. However, the constraint of the tamper ratio may be too stringent. This paper proposes a neighborhood characteristic based detection model for statistical fragile watermarking to lift the constraints of the tampered area from 4% to 14% of the host image. The neighborhood characteristic is used to effectively decrease the probability of false rejection, while the low probability of false acceptance is guaranteed by selecting proper threshold. The tamper detection performance of the proposed detection model is analytical analyzed in the different conditions. Analytical and experimental results demonstrate that the neighborhood characteristic based detection model effectively reduce the total number of false decisions and detect the tampered pixels with high probability.

In this paper, we propose a chaos and NDFT-based spread spectrum technique to conceal fingerprint-biometrics templates into audio signals. Fingerprint templates are encrypted by chaotic encryption, encoded by the BCH codes, modulated by chaotic parameter modulation (CPM), and then hid into the chaotically selected random sampling points of the host speech signal by a novel non-uniform discrete Fourier transform (NDFT)-based data hiding method. The template extraction process is completely blind and does not require original speech signal, thus the extraction depends on the secret key. Experimental and simulation results show that the proposed scheme is robust against common signal processing attacks, secure by secret keys, efficient in performance, and accomplishes perceptual transparency by exploiting the masking effects of human auditory system (HAS).

To overcome the computational complexity of the Volterra filter, a novel adaptive joint process filter using pipelined bilinear polynomial architecture (JPBPF) is proposed in this paper. The proposed architecture consists of two subsections: nonlinear subsection performing a nonlinear mapping from the input space to an intermediate space by the bilinear polynomial filter (BPF), and a linear filter performing a linear mapping from the intermediate space to the output space. The corresponding adaptive algorithms are deduced for the nonlinear subsection and linear filter subsection, respectively. To evaluate the performance of the JPBPF, a series of simulations are presented including nonlinear system identification, predicting of speech signals and nonlinear channel equalization. Compared with the conventional second-order Volterra (SOV) filter and BPF, the JPBPF exhibits a slightly better convergence performance in terms of convergence speed and steady-state error. Moreover, since those modules of a JPBPF can be performed simultaneously in a pipelined parallelism fashion, this would lead to a significant improvement in its total computational efficiency.

This paper proposes an adjacent-block based statistical detection method for self-embedding watermarking techniques to accurately identify the tampered blocks, and gives an analytical analysis of the tamper detection performance. In the proposed statistical detection method, we take all adjacent blocks of the test block and its mapping block into account and then utilize a statistic-based rule to verify the validity of image blocks. Analytical analysis and experimental results demonstrate that the proposed statistical detection method can identify the tampered blocks with a probability more than 98% even the tampered area is up to 70% of the host image. In addition, the proposed method outperforms conventional self-embedding fragile watermarking algorithms in tamper detection under collage attack and content-tampering attack.

A computationally efficient pipelined functional link artificial recurrent neural network (PFLARNN) is proposed for nonlinear dynamic system identification using a modification real-time recurrent learning (RTRL) algorithm in this paper. In contrast to a feedforward artificial neural network (such as a functional link artificial neural network (FLANN)), the proposed PFLARNN consists of a number of simple small-scale functional link artificial recurrent neural network (FLARNN) modules. Since those modules of PFLARNN can be performed simultaneously in a pipelined parallelism fashion, this would result in a significant improvement in its total computational efficiency. Moreover, nonlinearity of each module is introduced by enhancing the input pattern with nonlinear functional expansion. Therefore, the performance of the proposed filter can be further improved. Computer simulations demonstrate that with proper choice of functional expansion in the PFLARNN, this filter performs better than the FLANN and multilayer perceptron (MLP) for nonlinear dynamic system identification.

In this paper, we propose a novel method for image feature extraction, namely, (2D)^2PCALDA. This method directly extracts the optimal projective vectors from 2D image matrices by simultaneously considering row-direction 2DPCA and column-direction 2DLDA. The proposed method not only avoids huge feature matrix problem in 2DPCA and 2DLDA, but also take full advantage of the discriminant information and descriptive information of the images. Experiment results on Yale face database and ORL face databases demonstrate the effectiveness and robustness of the proposed method.

This paper proposes a novel computational efficient adaptive nonlinear equalizer based on combination of finite impulse response (FIR) filter and functional link artificial neural network (CFFLANN) to compensate linear and nonlinear distortions in nonlinear communication channel. This convex nonlinear combination results in improving the speed while retaining the lower steady-state error. In addition, since the CFFLANN needs not the hidden layers, which exist in conventional neural-network-based equalizers, it exhibits a simpler structure than the traditional neural networks (NNs) and can require less computational burden during the training mode. Moreover, appropriate adaptation algorithm for the proposed equalizer is derived by the modified least mean square (MLMS). Results obtained from the simulations clearly show that the proposed equalizer using the MLMS algorithm can availably eliminate various intensity linear and nonlinear distortions, and be provided with better anti-jamming performance. Furthermore, comparisons of the mean squared error (MSE), the bit error rate (BER), and the effect of eigenvalue ratio (EVR) of input correlation matrix are presented.

Non-intrusive methods for eye tracking are important for many applications of vision-based human computer interaction. However, due to the high nonlinearity of eye motion, how to ensure the robustness of external interference and accuracy of eye tracking poses the primary obstacle to the integration of eye movements into todays’s interfaces. In this paper, we present a strong tracking finite-difference extended Kalman filter algorithm, aiming to overcome the difficulty in modeling nonlinear eye tracking. In filtering calculation, strong tracking factor is introduced to modify a priori covariance matrix and improve the accuracy of the filter. The filter uses finite-difference method to calculate partial derivatives of nonlinear functions for eye tracking. The latest experimental results show the validity of our method for eye tracking under realistic conditions.

To address the problems of the inferior localization and high probability of false rejection in existing self-recovery fragile watermarking algorithms, this paper proposes a new self-recovery fragile watermarking scheme with superior localization, and further discusses the probability of false rejection (PFR) and the probability of false acceptance (PFA) of the proposed scheme. Moreover, four measurements are defined to evaluate the quality of a recovered image. In the proposed algorithm, the original image is divided into 2×2 blocks to improve localization precision and decrease PFR under occurrence of random tampering. The PFR under occurrence of region tampering can be effectively decreased by randomly embedding the water-mark of each block in conjunction with a novel method of tamper detection. Compared with the current self-recovery fragile watermarking algorithms, the proposed scheme not only resolves the tamper detection problem of self-embedding water-marking, but also improves the robustness against the random tampering of self-embedding watermarking. In addition, the subjective measurements are provided to evaluate the performance of the self-recovery watermarking schemes for image authentication.

This letter presents an efficient multimodal face and fingerprint biometrics authentication system on space-limited tokens, e.g. smart cards, driver license, and RFID cards. Fingerprint templates are encrypted and encoded/embedded into the face images in such a way that the features, which are used in face matching, are not significantly changed during encoding and decoding. As a result, the verification accuracy based on decoded images is similar to that with original images. Experimental and simulation results show that the proposed scheme is an efficient and a cheap solution to the multimodal biometrics authentication on space-limited tokens without degrading the overall decoding and matching performance of the biometrics system. Besides, the proposed system can also be beneficial to cope with the problems of unimodal biometrics systems.

The drawbacks of the current authentication watermarking schemes for JPEG images, which are inferior localization and the security flaws, are firstly analyzed in this paper. Then, two counterferiting attacks are conducted on them. To overcome these drawbacks, a new digital authentication watermarking scheme for JPEG images with superior localization and security is proposed. Moreover, the probabilities of tamper detection and false detection are deduced under region tampering and collage attack separately. For each image block, the proposed scheme keeps four middle frequency points fixed to embed the watermark, and utilizes the rest of the DCT coefficients to generate 4 bits of watermark information. During the embedding process, each watermark bit is embedded in another image block that is selected by its corresponding secret key. Since four blocks are randomly selected for the watermark embedding of each block, the non-deterministic dependence among the image blocks is established so as to resist collage attack completely. At the receiver, according to judging of the extracted 4 bits of watermark information and the corresponding 9-neighbourhood system, the proposed scheme could discriminate whether the image block is tampered or not. Owing to the diminishing of false detection and the holding of tamper detection, we improve the accuracy of localization in the authentication process. Theoretic analysis and simulation results have proved that the proposed algorithm not only has superior localization, but also enhances the systematic security obviously.

Conventional discriminant locality preserving projection (DLPP) is a dimensionality reduction technique based on manifold learning, which has demonstrated good performance in pattern recognition. However, because its objective function is based on the distance criterion using L2-norm, conventional DLPP is not robust to outliers which are present in many applications. This paper proposes an effective and robust DLPP version based on L1-norm maximization, which learns a set of local optimal projection vectors by maximizing the ratio of the L1-norm-based locality preserving between-class dispersion and the L1-norm-based locality preserving within-class dispersion. The proposed method is proven to be feasible and also robust to outliers while overcoming the small sample size problem. The experimental results on artificial datasets, Binary Alphadigits dataset, FERET face dataset and PolyU palmprint dataset have demonstrated the effectiveness of the proposed method.

In this paper, we propose a matrix random low-rank approximation (MRLRA) approach to generate cancelable biometric templates for privacy-preserving. MRLRA constructs a random low-rank matrix to approximate the hybridization of biometric feature and a random matrix. Theoretically analysis shows the distance between one cancelable low-rank biometric template by MRLRA and its original template is very small, which results to the verification and authentication performance by MRLRA is near that of original templates. Cancelable biometric templates by MRLRA conquer the weakness of random projection based cancelable biometric templates, in which the performance will deteriorate much under the same tokens. Experiments have verified that (i) cancelable biometric templates by MRLRA are sensitive to the user-specific tokens which are used for constructing the random matrix in MRLRA; (ii) MRLRA can reduce the noise of biometric templates; (iii)Even under the condition of same tokens, the performance of cancelable biometric templates by MRLRA doesn't deteriorate much.

Linear discriminant analysis (LDA) is a well-known dimensionality reduction technique, which is widely used for many purposes. However, conventional LDA is sensitive to outliers because its objective function is based on the distance criterion using L2-norm. This paper proposes a simple but effective robust LDA version based on L1-norm maximization, which learns a set of local optimal projection vectors by maximizing the ratio of the L1-norm-based between-class dispersion and the L1-norm-based within-class dispersion. The proposed method is theoretically proved to be feasible and robust to outliers while overcoming the singular problem of the within-class scatter matrix for conventional LDA. Experiments on artificial datasets, standard classification datasets and three popular image databases demonstrate the efficacy of the proposed method.

Centre-symmetric local derivative pattern CS-LDP algorithm is proposed to describe the local second-order derivative feature of texture. However, CS-LDP can only describe second-order derivative feature of texture on four directions and lost some discriminant information on other directions. Addressing such problems, this paper proposed multi-resolution elongated CS-LDP ME-CS-LDP to solve such problem. By increasing the number of directions, which can be implemented by increasing the sampling points on the ellipse radius with interpolation, multi-resolution elongated CS-LDP can provide more discriminant information on more directions. Furthermore, our proposed multi-resolution elongated CS-LDP is defined in ellipse field to depict some important ellipse part of faces, like eyes and mouth. Gabor filter plus ME-CS-LDP/weighed ME-CS-LDP is used for face recognition in this paper. The experiment results on the illumination subset of Yale B database, the subset of PIE illumination database and VALID face database have validated the effectiveness of the proposed method.

The security and privacy of biometric template has become a hot issue with the development of biometric technology. Palmprint verification has a high accuracy thank to texture coding methods, so it is urgent to develop nice cancelable palmprint coding scheme based on texture feature. PalmHash Code is proposed as a secure and effective cancelable palmprint coding framework in this paper. Three generation mechanisms of PalmHash Code are introduced and compared. Beside, two measures, perpendicular orientation transposition and multi-orientation score level fusion, are used to improve the performance of PalmHash Codes. The experimental results and analysis confirm that PalmHash Code enhanced by the two measures meet the four objectives of cancelable biometric satisfactorily.