Search

This paper presents an isolated sign language recognition system that comprises of two main phases: hand tracking and hand representation. In the hand tracking phase, an annotated hand dataset is used to extract the hand patches to pre-train Convolutional Neural Network (CNN) hand models. The hand tracking is performed by the particle filter that combines hand motion and CNN pre-trained hand models into a joint likelihood observation model. The predicted hand position corresponds to the location of the particle with the highest joint likelihood. Based on the predicted hand position, a square hand region centered around the predicted position is segmented and serves as the input to the hand representation phase. In the hand representation phase, a compact hand representation is computed by averaging the segmented hand regions. The obtained hand representation is referred to as “Hand Energy Image (HEI)”. Quantitative and qualitative analysis show that the proposed hand tracking method is able to predict the hand positions that are closer to the ground truth. Similarly, the proposed HEI hand representation outperforms other methods in the isolated sign language recognition.

The flight path of the unmanned aerial vehicle (UAV) relay can significantly affect the performance of the wireless relay communication system. In this paper, we investigate the optimization of the UAV flight path in a wireless relay communication system, when a jamming or interference signal is present. The UAV flight path is optimized to maximize the signal-to-interference-plus-noise ratio (SINR), and the required parameters are estimated. The simulation results show that the proposed method can achieve performance close to an ideal SINR optimization and also highlight the benefits of optimizing the SINR instead of the signal-to-noise ratio (SNR) when a jamming or interference signal is present.

In this paper, we propose a hand tracking method which was inspired by the notion of the four dukkha: birth, aging, sickness and death (BASD) in Buddhism. Based on this philosophy, we formalize the hand tracking problem in the BASD framework, and apply it to hand track hand gestures in isolated sign language videos. The proposed BASD method is a novel nature-inspired computational intelligence method which is able to handle complex real-world tracking problem. The proposed BASD framework operates in a manner similar to a standard state-space model, but maintains multiple hypotheses and integrates hypothesis update and propagation mechanisms that resemble the effect of BASD. The survival of the hypothesis relies upon the strength, aging and sickness of existing hypotheses, and new hypotheses are birthed by the fittest pairs of parent hypotheses. These properties resolve the sample impoverishment problem of the particle filter. The estimated hand trajectories show promising results for the American sign language.

Gait has unique advantage at a distance when other biometrics cannot be used since they are at too low resolution or obscured, as commonly observed in visual surveillance systems. This paper provides a survey of the technical advancements in vision-based gait recognition. A wide range of publications are discussed in this survey embracing different perspectives of the research in this area, including gait feature extraction, classification schemes and standard gait databases. There are two major groups of the state-of-the-art techniques in characterizing gait: Model-based and motion-free. The model-based approach obtains a set of body or motion parameters via human body or motion modeling. The model-free approach, on the other hand, derives a description of the motion without assuming any model. Each major category is further organized into several subcategories based on the nature of gait representation. In addition, some widely used classification schemes and benchmark databases for evaluating performance are also discussed.

A normalized histogram of optical flow as a hand representation of the sign language.Block-based histogram provides spatial information and local translation invariant.Block-based histogram of optical flow enables sign language length invariance. In this paper, we propose a block-based histogram of optical flow (BHOF) to generate hand representation in sign language recognition. Optical flow of the sign language video is computed in a region centered around the location of the detected hand position. The hand patches of optical flow are segmented into M spatial blocks, where each block is a cuboid of a segment of a frame across the entire sign gesture video. The histogram of each block is then computed and normalized by its sum. The feature vector of all blocks are then concatenated as the BHOF sign gesture representation. The proposed method provides a compact scale-invariant representation of the sign language. Furthermore, block-based histogram encodes spatial information and provides local translation invariance in the extracted optical flow. Additionally, the proposed BHOF also introduces sign language length invariancy into its representation, and thereby, produce promising recognition rate in signer independent problems.

Signal-to-noise ratio (SNR) estimation available in the literature are designed based on the assumption of Gaussian noise models. These estimators may produce misleading results when the distribution of the noise deviates from Gaussian. This paper investigates the performance of existing SNR estimators in an additive non-Gaussian noise channel based on a Gaussian mixture model. An expectation---maximization (EM) based approach is proposed for optimum SNR estimation in the non-Gaussian noise channel. In addition, the Cramer---Rao bound is derived and used as a benchmark to assess the performance of the SNR estimators. Simulation results confirm the optimality and robustness of the proposed EM-based estimator in Gaussian and non-Gaussian noise channels.

A fusion of median and mode filtering for better background model.A serial particle filter that can better detect and track the object of interest.A novel covariance matrix feature for isolated sign language representation. As is widely recognized, sign language recognition is a very challenging visual recognition problem. In this paper, we propose a feature covariance matrix based serial particle filter for isolated sign language recognition. At the preprocessing stage, the fusion of the median and mode filters is employed to extract the foreground and thereby enhances hand detection. We propose to serially track the hands of the signer, as opposed to tracking both hands at the same time, to reduce the misdirection of target objects. Subsequently, the region around the tracked hands is extracted to generate the feature covariance matrix as a compact representation of the tracked hand gesture, and thereby reduce the dimensionality of the features. In addition, the proposed feature covariance matrix is able to adapt to new signs due to its ability to integrate multiple correlated features in a natural way, without any retraining process. The experimental results show that the hand trajectories as obtained through the proposed serial hand tracking are closer to the ground truth. The sign gesture recognition based on the proposed methods yields a 87.33% recognition rate for the American Sign Language. The proposed hand tracking and feature extraction methodology is an important milestone in the development of expert systems designed for sign language recognition, such as automated sign language translation systems.

Eavesdropping occurs when an unauthorized receiver secretly intercepts a transmitted signal. In this paper, we investigate the optimum flight path and transmit power for an unmanned aerial vehicle (UAV) relay to improve secrecy performance in the presence of an eavesdropper. We propose an algorithm based on particle swarm optimization (PSO) where the UAV flight path and power are jointly optimized to maximize the secrecy capacity. Simulation results show that the proposed algorithm is capable of achieving a performance that is close to the ideal secrecy performance. Simulation results also highlight the benefits of power optimization.

Transient Binary Patterns of gait sequence.Inherently combines both appearance information and temporal information.Less sensitive to silhouette noise in individual frames. In this work, we present a combination of spatiotemporal approach and texture descriptors to extract the temporal patterns in gait cycles. Unlike most conventional methods that focus on spatial information while limiting temporal information captured, spatiotemporal methods preserve both spatial and temporal information. Inspired by the success of texture descriptors in face recognition, the proposed method likewise constructs texture descriptors of gait motion over time. For each gait cycle, the pixel-wise binary patterns along the temporal axis, referred to as the Transient Binary Patterns (TBP), is analyzed. These pixel-wise TBPs are then grouped into regional blocks from which we construct regional TBP histograms. These regional TBP histograms collectively form the global TBP histogram that represents both the distribution of temporal patterns and spatial location. Experimental results clearly show the superiority of the proposed approach over other considered methods.

This paper presents a gait recognition method using the combination of motion history image (MHI), binarized statistical image features (BSIF) and histograms of oriented gradients (HOG). The method first encodes the motion pattern and direction of the gait cycle in motion history image. Subsequently, performing convolution on the motion history image using pre-learnt filters as kernel, binarized statistical image features are generated by summing the convolution output images. Histograms of oriented gradients are then computed on binarized statistical image features. Gait signature of a gait cycle is attained by accumulating all the HOG descriptors. Experimental result shows that the proposed method performs promisingly in gait recognition.

Voice activity detection (VAD) is a vital process in voice communication systems to avoid unnecessary coding and transmission of noise. Most of the existing VAD algorithms continue to suffer high false alarm rates and low sensitivity when the signal-to-noise ratio (SNR) is low, at 0 dB and below. Others are developed to operate in offline mode or are impractical for implementation in actual devices due to high computational complexity. This paper proposes the upper envelope weighted entropy (UEWE) measure as a means to enable high separation of speech and non-speech segments in voice communication. The asymmetric nonlinear filter (ANF) is employed in UEWE to extract the adaptive weight factor that is subsequently used to compensate the noise effect. In addition, this paper also introduces a dual-rate adaptive nonlinear filter (DANF) with high adaptivity to rapid time-varying noise for computation of the decision threshold. Performance comparison with standard and recent VADs shows that the proposed algorithm is superior especially in real-time practical applications.

In this paper, a real-time robust voice activity detector (VAD) is proposed. The proposed VAD adopts the gammatone filter and modify the existing long-term signal variability (LTSV) measure, i.e. known as GMLTSV in short. The proposed VAD is an improved version of existing VAD which used gammatone filter and entropy. The LTSV measure is modified to adapt to the amplitude envelopes extracted using gammatone filter by swapping entropy and variance used in LTSV measure to reduce noise effect in the extracted temporal envelopes and improve discriminative power of the extracted feature. The proposed algorithm also implements an adaptive threshold that is computed using a nonlinear filter to track short-term trend of the extracted feature in real-time. The proposed VAD using GMLTSV feature is tested against clean speech signals from TIMIT test corpus which are degraded at SNR ranged from -10dB to 20dB by non-stationary noise, eg. airport noise, babble noise, exhibition noise from Aurora-2 database, and stationary noise, eg. additive white Gaussian noise. Based on the evaluation, it is proven that the proposed GMLTSV-based VAD is robust in speech and non-speech detection even at low signal-to-noise ratio (SNR) and outperformed other existing voice activity detectors which are compared in the evaluation. The proposed VAD achieved satisfactory accuracy when compared to the impractical single frequency filtering based VAD while implementing real-time scheme for practical application.

In this paper, a Temporal Gradient Patterns (TGP) method is proposed for gait recognition. The method first computes the gradients of the silhouette in each image. Subsequently, the gradient of each pixel determines the bin number to which the pixel belongs to. The bin number of current and next frame jointly cast a vote to the corresponding index in the matrix of oriented gradients. The obtained matrix henceforth encodes the gradient pattern of each pixel in the gait cycle. The TGP method not only describes the spatial silhouette shapes but also implicitly captures the silhouette deformation in temporal axis. Experimental results show that the proposed approach attains a promising recognition rates.

In this paper, we propose a new probabilistic gait representation to characterize human walking for recognition by gait. The approach obtains the binomial distribution of every pixel in a gait cycle. Organizing the binomial distribution of all pixels in the gait image, we obtain the gait signature, which we denote as the Gait Probability Image (GPI). In the recognition stage, symmetric Kullback-Leibler divergence is used to measure the information theoretical distance between gait signatures. The experimental results reveal that GPI achieves promising recognition rates. Besides that, experiments on different walking speeds demonstrate that GPI is robust to slight variation in walking speed.

In this paper, we propose a time-sliced averaged motion history image (TAMHI) alongside the histograms of oriented gradients (HOG) to generate gait signatures in a gait recognition problem. Building on the motion history image (MHI), TAMHI divides the gait cycle into several regular time windows to generate the same number of TAMHI composite images. HOG descriptors are then calculated on these composite images to obtain the gait signature. The time-slicing procedure to produce multi-composite images preserve more detailed transient information of gait cycles. Additionally, time-normalization also introduces gait length invariancy into the representation, hence, offering a better recognition rate to slight changes in walking speed.

In this letter, a jamming resilient communication technique for unmanned aircraft vehicle (UAV) is proposed. Multiple-input multiple-output (MIMO) interference cancellation is exploited to combat reactive jamming. An improved channel estimation algorithm based on Kalman filter and basis expansion model (BEM) is used to track the quick changes of the combined channel components. Simulation results show that the performance of the proposed technique is significantly better than the existing jamming resilient communication technique.

Voice activity detector (VAD) is used to detect the presence or absence of human voice in a signal. A robust VAD algorithm is essential to distinguish human voice in a noisy acoustic signal. There were many recent works in development of robust VAD which focus on unsupervised features extraction such as temporal variation, signal-to-noise ratio in [1] and etc. However, these methods are typically sensitive to nonstationary noise especially under low SNR. To overcome these problems, this paper presents a robust voice activity detection (VAD) method via a combination of gammatone filtering and entropy as an information-theoretic measure in the detection algorithm. The performance of the proposed algorithm is tested using speech signals from TIMIT test corpus with additive noise at varying degrees of signal-to-noise ratio. The results show that the proposed robust VAD outperforms other existing methods in terms of detection accuracy.

Finger spelling is a way of communication by expressing words using hand signs in order to ensure deaf and dumb community can communicate with others effectively. Therefore, a system that can understand finger spelling is needed. As a result of that, this work is conducted to primarily develop a tutoring system for finger spelling. To develop a robust real-time finger spelling tutoring system, it is necessary to ensure the accuracy of the finger spelling recognition. Even though there are existing solutions available for a decade, but most of them are just focusing on improving accuracy rate without implementing their solutions as a complete tutoring system for finger spelling. Consequently, it inspires this research project to develop a tutoring system for finger spelling. Microsoft Kinect sensor is used to acquire color images and depth images of the finger spells. Depth images are used to perform segmentation on the color images. After that, the segmented images are used as input and pass into a two hidden layers backpropagation neural network for classification.

A novel competitive approach to particle swarm optimization (PSO) algorithms is proposed in this paper. The proposed method uses extrapolation technique with PSO (ePSO) for solving optimization problems. By considering the basics of the PSO algorithm, the current particle position is updated by extrapolating the global best particle position and the current particle positions in the search space. The position equation is formulated with the global best (gbest) position, local best position (pbest) and the current position of the particle. The proposed method is tested with a set of 13 standard optimization benchmark problems and the results are compared with those obtained through two existing PSO algorithms, the canonical PSO (cPSO), the Global-Local best PSO (GLBest PSO). The cPSO includes a time-varying inertia weight (TVIW) and time-varying acceleration co-efficients (TVAC) while the GLBest PSO consists of Global-Local best inertia weight (GLBest IW) with Global-Local best acceleration co-efficient (GLBestAC). The simulation results clearly elucidate that the proposed method produces the near global optimal solution. It is also observed from the comparison of the proposed method with cPSO and GLBest PSO, the ePSO is capable of producing a quality of optimal solution with faster convergence rate. To strengthen the comparison and prove the efficacy of the proposed method a real time application of steel annealing processing (SAP) is also considered. The optimal control objectives of SAP are computed through the above said three PSO algorithms and also through two versions of genetic algorithms (GA), namely, real coded genetic algorithm (RCGA) and hybrid real coded genetic algorithm (HRCGA) and the results are analyzed with the proposed method. From the results obtained through benchmark problems and the real time application of SAP, it is clearly seen that the proposed ePSO method is competitive to the existing PSO algorithms and also to GAs.

Recently, a speech model inspired by signal subspace methods was proposed for a speech classifier. In using subspace information to characterize the speech signal, subspace trajectories in the form of the right singular vectors of the measurement matrices are obtained. Signal classification is thereafter accomplished by a minimum-distance rule with noteworthy results. This paper extends the foregoing approach by organizing the vector trajectories into matrices. The matrices so obtained are the reduced-rank approximation of the sample correlation matrices. A new dissimilarity measure in the Frobenius norm is correspondingly proposed for the matrix trajectories. Simulation results of the proposed composite signal subspace classifier in an isolated digit speech recognition problem reveal an improved performance over its predecessor. Additionally, the results also show the proposed classifier retaining the white noise robustness of the original design.

A speech model inspired by the signal subspace approach was recently proposed as a speech classifier with modest results. The method entails, in general, the assemblage of a set of subspace trajectories that consist of the right singular vectors of measurement matrices of the signal under consideration. Given an unknown signal, a simple distortion measure then applies in the classification procedure to pick the best matched class prototype. This letter examines the issue of robustness in the subspace classification scheme. Borrowing an important result on noisy measurement matrices, this letter formally establishes the notion of robustness in subspace classification and proceeds to propose a class of robust distortion measures for signal subspace models. Simulation results of subspace classifiers implementing the new distortion measures in an isolated digit speech recognition problem reveal no degradation in recognition accuracy, even under low SNR conditions.

In this paper, a speech classifier inspired by the signal subspace approach is developed. A novel signal subspace speech model is initially obtained via a rank reducing subspace decomposition algorithm that is based on the SVD. Motivated by the assumption that the speech signal comprises of short term dynamics that are slowly changing, it follows that the signal subspace of the speech signal is likewise slowly changing. The proposed signal subspace model aims to characterize the subspace dynamics using a family of subspace trajectories. In particular, each subspace trajectory is a sequence of vectors that traces the dynamics of a rank-one subspace in time. An assembly of these trajectories, henceforth, specifies the progression of the embedded signal subspace. To construct the signal subspace classifier, prototype elements in the form of the signal subspace models are determined for every signal class. A minimum-distance rule with a distance measure that resembles an energy difference function is subsequently applied in the actual classification task. Simulation of the proposed signal subspace classifier in an isolated digit speech recognition problem reveals promising results.

A speech model inspired by the signal subspace methods was recently proposed as a speech classifier with modest results. Fashioned along a "best representation" approach, the absence of valuable interclass information in the speech model, however, impairs the ability of the classifier to distinguish between phonetically alike classes. This letter proposes an improved classifier that implements interclass information. Specifically, a measure of the discriminative quality of individual class elements is defined and determined for all class elements. The discrimination measures thus obtained are subsequently applied in the classification procedure. Simulation results of the proposed signal subspace classifier in an isolated digit speech recognition problem reveal an improved performance over its predecessor

The RF subsampling receiver suffers from nonlinear distortion caused by the front end of the receiver, specifically the low noise amplifier. Strong interferers present in the receiver introduce cross modulation distortion to the desired signal. This paper introduces a compensation method for the cross modulation distortion using digital signal processing techniques. In the proposed method, two signals are down converted using the subsampling receiver and digitized. A joint channel and nonlinearity estimation is then performed for the mitigation of the cross modulation distortion. Simulation results show that the proposed method obtains more accurate estimates compared to existing work, and subsequent compensation is improved as shown by the recovery of the desired signal.

In this paper, a robust joint carrier frequency offset (CFO) and fast time-varying channel tracking algorithm for MIMOOFDM systems is proposed. The proposed method is robust to noise statistics and noise distribution uncertainties. The extended H  filter (EHF) is employed for the joint tracking process as it does not require any knowledge of noise statistics or noise distribution, unlike the conventional extended Kalman filter (EKF). To reduce computational complexity, the tracking of fast time-varying channel gains is facilitated by the basis expansion model (BEM). Simulation results show that the performance of the proposed algorithm without knowledge of noise statistics matches the performance of the existing algorithm under Gaussian noise and is better than that of the existing algorithm under non-Gaussian noise.

This paper presents a quantization approach to the detection of singular points in fingerprint images. Singular points in fingerprint images referred to the core and delta points. They are the global features found in the orientation field image of a fingerprint. A singular point is normally detected at the intersection of three gray levels of a quantized orientation field image at quantization level-3 but no study has been carried out for higher quantization level. In this paper, the singular point detection has been investigated up to quantization level-8 by employing two performance measures: sensitivity and precision. The performance at level-5 has been found to be optimal when both measures are combined.

This paper presents a design approach to determine the subsampling frequency ranges for a multi band system, where only one signal is prioritized so that it doesn't overlap with the other signals during the subsampling process. The proposed technique will introduce the formulas derived from an existing method to relax the conditions on the other signals in the multi band case. The results show that by having this assumption, the subsampling frequency needed to extract the desired signal is lower than the previously derived methods.

This article introduces a new multiuser detection scheme which uses evolutionary programming (EP) to detect the user bits based on the maximum-likelihood decision rule. The major advantage of the proposed detector is that it has a lower computational complexity compared to other popular evolutionary-algorithm-based detectors. The simulation results show that the EP has always converged to the optimum solution with a small number of generations. The simulated average computational time performance demonstrates that this approach achieves practical ML performance with polynomial complexity in the number of users.

The use of constant modulus algorithm (CMA) and correntropy in blind adaptive equalization requires choosing a suitable cost function such that the FIR equalizer removes sufficient intersymbol interference (ISI). A correntropy cost function based on minimizing the absolute difference between the correntropy of equalizer output and source symbol is proposed and compared with conventional CMA and correntropy cost functions. Simulation study on both Gaussian and impulsive additive noise scenarios show improved overall performance.

Joint channel and frequency offset tracking based on an autoregressive (AR) model are often performed by either assuming the knowledge of the state transition matrix and state noise or obtaining the state transition matrix and state noise via the Yule Walker equations. In this paper, we propose a joint carrier frequency offset (CFO) and channel parameters tracking algorithm for MIMO-OFDM system which is capable of tracking the channel parameters without requirements of additional information such as maximum Doppler frequency and maximum delay spread. The Extended Kalman Filter is proposed for the joint tracking process. Simulation results show that the performance of the proposed method is close to the case when channel parameters are known. (6 pages)

Most of the signal-to-noise ratio (SNR) estimators published in literature are designed based on Gaussian noise assumption. These estimation schemes typically perform poorly when the additive noise has a non-Gaussian distribution. This paper investigates the robustness of several popular SNR estimators in two-term Gaussian mixture noise. The Cramer-Rao bound is derived and used as a benchmark against which the performance of the estimators is measured. Simulations results show that the SNR estimators suffer performance degradation in non-Gaussian noise channels.

This project aims to enhance human speech energy at low frequencies using linear prediction methods to achieve better performance for speech recognizers in noisy conditions. In order to achieve this, a recognition system based on Warped Linear Prediction (WLP) is proposed. WLP is based on Warped Fourier Transform and the consideration is to warp a signal to another frequency scale and perform Fourier Transform on the warped scale. This technique will transform speech signal so that the frequency resolution at the lower frequency region is higher, thus more detailed information on the signal can be obtained from the low frequencies. After the signal is transformed through warping, cepstral coefficients are obtained and it can be acknowledged as Warped Linear Prediction. Evaluation of the effectiveness of this method has been conducted in isolated word recognition tests. Experimental results show that the WLP performs better than linear prediction method for the set SNR range, based on two distortion measures that were tried. The new method shows no degradation in recognition accuracy under high SNR conditions, but performs significantly better under low SNR conditions. At SNR of 4dB, performance improvements of up to 70 percent can be seen.

Adopting frequency-domain equalization (FDE) in DS-CDMA can partially restore the orthogonality of users and suppress multiple-access interference (MAI). The existing FDEs are designed with the assumption of additive white Gaussian noise. Although the assumption is quite proper for a variety of applications, it has been revealed that many background noises arising in the reality are with non-Gaussian statistics. In this paper, we show that the Gaussian-based FDEs fare poorly in non-Gaussian noise environment. To achieve reliable data transmission in frequency-selective multiple-access channel with unknown non-Gaussian noise, we propose a robust adaptive FDE which consists of a nonlinear preprocessing front-end followed by a linear adaptive FDE. The nonlinear front-end adapts itself to the unknown non-Gaussian noise to suppress the impulses while the FDE effectively removes the MAI. Results from the simulation reveal that the proposed robust FDE is able to perform successfully in both Gaussian and non-Gaussian noise environments.