Energy efficiency (EE) optimization of joint power allocation (PA) and beamforming (BF) is studied for multiuser distributed multiple-input multiple-output (D-MIMO) systems with maximum ratio combining. Subject to power budget constraints, a non-convex constrained optimization problem is firstly formulated to maximize EE of single-user D-MIMO. We propose a near-optimal joint design scheme with the block coordinate descent method, which has an effective iterative algorithm to find optimal BF for fixed PA, and an iterative algorithm based on the Dinkelbach methods to find optimal PA for fixed BF. We also develop a low-complexity suboptimal joint scheme, which has closed-form BF and PA by maximizing an effective signal-to-noise ratio for fixed PA and a derived EE upper bound for fixed BF, respectively. It has lower complexity than the near-optimal scheme with slight performance loss. With these results, two joint design schemes are developed for multiuser D-MIMO. The first scheme can achieve the best performance, while the second one based on closed-form PA and BF has slight performance loss in comparing to that of the former, but with lower complexity. Simulation results verify the effectiveness of the developed schemes over some existing schemes.

In this paper, the error performance and spectrum efficiency (SE) of uplink massive multiuser spatial modulation (SM) MIMO systems with zero-forcing detection for perfect and imperfect channel state information (CSI) are investigated over composite Rayleigh fading channels. We firstly derive the error probabilities of transmit antenna index detection and transmitted symbol detection of each user. Using these results, the bit error rate (BER) performance of the system under perfect CSI is analyzed. Furthermore, a closed-form average BER of the system is derived. We generalize the analytical results of perfect CSI to those of imperfect CSI for practicality, and derive a closed-form BER for the system with imperfect CSI. Based on this, two asymptotic BER expressions are developed to characterize the error performance of the system for large-scale receive antennas and high SNR. By means of the asymptotic expressions, the diversity order of the system is attained. Besides, we derive a lower bound on the achievable SE of the system with imperfect CSI. With this lower bound, an approximate expression of the achievable SE is obtained in closed form. Simulation results indicate that our theoretical analysis is valid, and agrees with simulation results well.

In this paper, we develop an optimal precoder and a suboptimal precoder for uplink distributed antenna systems with transmit correlation over Rician channels. The objective is to minimize a pairwise error probability (PEP) with constrained transmit power. The optimal scheme uses an efficient projection-based steepest descent algorithm to obtain the Gram matrix of the precoder, while the suboptimal scheme exploits the eigen-structure of the Gram matrix to obtain its unitary matrix (beamforming) and eigenvalues (power control). For a given beamforming (BF), an efficient power adaptation based on a fuzzy signal-to-noise ratio (SNR) is developed. The PEP analysis at low SNR is used to obtain a suboptimal BF. The derived BF and the fuzzy power adaptation have closed-form expressions that make the suboptimal scheme computationally efficient in comparison with existing methods. The BF can be re-optimized via the Riemannian steepest descent method to further reduce the PEP. Simulation results show that the proposed optimal scheme not only gives the performance the same as the existing optimal scheme when the latter can converge, but it also has lower computational complexity and faster convergence. Moreover, the suboptimal scheme can give a nearly optimal error rate.

In this study, an analytical analysis of normalised spatial correlation based on a hierarchical angle model for uniform linear array (ULA) and uniform circular array (UCA) over clustered narrowband multiple‐input multiple‐output channels is presented. The clustered channel has independent multiple paths, each of which is a cluster consisting of a number of independent subpaths, constructing a hierarchical angle structure. Unlike most of existing analysis methods that employ various simplification to the hierarchical angle, in this study, the hierarchical angle is expressed as a function of bivariate random variables. This formulation can include the simplified models as special cases. Assuming the angle of departure and angle of arrival of each subpath are independent, a closed‐form spatial correlation expression between transmit and receive antennas of ULA or UCA is derived. Simulation results show that the theoretical spatial correlation matches well with the simulated one with channel data generated from the 3GPP spatial channel model for different scenarios.

In this study, a joint design of precoder and equaliser of a linear transceiver for multiple‐input–multiple‐output system with orthogonal frequency division multiplexing in the presence of intercarrier interference (ICI) is presented. The matrix structures of the precoder and equaliser are banded for the sake of reducing the computational complexity and feedback overhead from the receiver to the transmitter. The design criterion is to minimise the mean‐squared error subject to a total transmitted power constraint of which the power is allocated over space and frequency domains in the precoder. The authors use the Karush–Kuhn–Tucker conditions to derive an iterative procedure to obtain a convergent solution and a closed‐form procedure for the optimal full transceiver. Numerical results show that the banded precoder is an efficient scheme to improve the bit error rate (BER) of the transceiver by simply increasing its band size and can provide better BER performance than that of the existing jointly designed full transceiver in the presence of ICI. With small band sizes, the proposed transceiver can give performance close to that of the jointly designed full transceiver but with lower implementation complexity and feedback overhead to the transmitter.

In this paper, unified BER performance and outage probability analyses of downlink multiuser scheduling for multiple-input multiple-output (MUS-MIMO) transmission/combining (transceiving) systems with imperfect channel state information (CSI) caused by feedback over Rayleigh fading channels is presented. The unified analysis is applicable to MIMO transceiving systems whose received signal-to-noise ratio (SNR) is given as largest order statistic of chi-square distribution. The performance of MUS-MIMO employing selective transmission/selective combining (ST/SC), ST/maximum ratio combining (MRC) and space-time block codes (STBC) can be analyzed by this general framework. Accurate and approximate unified closed-form BER and outage probability expressions of the MUS-MIMO transceiving system with normalized SNR-based scheduling in heterogeneous networks are derived. Based on the approximate BER and outage probability expressions, diversity gains are analyzed at high SNR. The results indicate that the MUS-MIMO for using the three transceiving schemes can achieve full diversity order for perfect CSI, while the diversity gain is degraded differently for the three transceiving schemes under imperfect CSI. Computer simulation shows that the theoretical analysis is in good agreement with simulation results, and the approximate BER expressions are very close to the accurate ones but with lower computational complexity.

In this paper, a statistically robust precoder design is presented for space-time coded multiple-input, multiple-output (MIMO) relay networks over double correlated Rician channels. We assume both the source and destination terminals are equipped with multiple-antenna, while a single antenna is used in distributed non-regenerative relay nodes for reducing feedback overhead and processing burden on relays. The precoding matrix at the source terminal and amplifying parameters of relay nodes are optimized to maximize the total signal-to-noise ratio (SNR) at the destination assuming only statistical channel state information (CSI) available at the transmission sides. The optimization problem can be efficiently solved by an iterative method, resulting in a single-beam mode source precoder and diagonal precoding matrix for relay nodes. Numerical results demonstrate that the proposed scheme outperforms the equal power allocation scheme and existing method with single antenna nodes.

In this paper, an analytical expression of normalized spatial correlation of a hierarchical angle structure for 3GPP Spatial Channel Model (SCM) is derived. Most of the existing formulas of the normalized spatial correlation were derived by simplifying the angle structure. In the paper, the hierarchical angle is considered as a function of multiple random variables, which can include those simplified angle structures as special cases. A closed-form expression of the probability density function of the hierarchical angle for uniform linear array is developed that can yield the analytical formula of normalized spatial correlation. Computer evaluation shows that the derived formula matches well with the simulated correlations generated from the 3GPP SCM.

The performance analysis of a space‐time coded multiple‐input multiple‐output (MIMO) system with variable‐rate adaptive modulation over flat Rayleigh fading channels for both perfect and imperfect channel state information (CSI) is presented. In this study, the optimum fading gain switching thresholds for attaining maximum spectrum efficiency (SE) subject to an average bit error rate (BER) constraint are derived. The existence and uniqueness of the Lagrange multiplier in the constrained SE optimisation is studied. It is shown that the Lagrange multiplier does exist and is unique for imperfect CSI. On the other hand, the Lagrange multiplier will be unique if the existence condition for MIMO under perfect CSI is satisfied. A practical iterative algorithm based on Newton's method for finding the Lagrange multiplier is proposed. By the switching thresholds, closed‐form expressions of the SE and average BER are obtained. Simulation results for SE and BER are in good agreement with the theoretical analysis. The results show that the space‐time block coded MIMO system using adaptive modulation (AM‐STBC‐MIMO) with average BER constraint provides SE better than AM‐STBC‐MIMO with fixed thresholds, and AM‐STBC‐MIMO using a BER upper bound, but it has performance degradation in SE for imperfect CSI.

The performance analysis of a space–time block coded multiple input multiple output system with variable-power (VP) adaptive modulation (AM) over Rician fading channels for imperfect channel state information is presented. The optimum switching thresholds for attaining maximum spectrum efficiency (SE) subject to a target bit error rate (BER) and an average power constraint are derived. In the derivation, a very tight BER expression of quadrature amplitude modulation (QAM) is used to develop a new power adaptation scheme that can fulfil the target BER even at low signal-to-noise ratio (SNR) and achieve higher SE than the scheme using the BER upper bound commonly adopted in the literature. The existence and uniqueness of the Lagrange multiplier used in the constrained optimisation are investigated. It is shown that the Lagrange multiplier is unique when it exists. By using the switching thresholds, the authors obtain closed-form expressions of SE and average BER of the system. Computer simulation shows that the theoretical analysis is in good agreement with simulation. The results show that the proposed VP-AM scheme provides better SE than the constant-power counterparts and the VP-AM scheme using the commonly used BER bound.

In this study, a precoding scheme for minimising bit error rate (BER) subject to a power constraint for space-time-coded multiple-input–multiple-output systems over flat Rayleigh fading channel in the presence of channel estimation error and feedback delay is presented, it includes the precoding with estimation error or feedback delay only as its special cases. With this scheme, an algorithm for attaining the optimal power allocation across multiple beams is developed. It is shown that the Lagrange multiplier for the constrained optimisation does exist and is unique. A practical iterative algorithm based on Newton's method for finding the Lagrange multiplier is proposed. Generally, the optimal power control may result in negative power allocation to some beams. A necessary and sufficient condition for the beams having positive power control is developed that simplifies the optimisation procedure. Simulation results show that the developed precoding scheme can provide BER lower than the equal power allocation scheme and the existing precoding scheme, but has performance degradation when compared with the precoding scheme with estimation error or feedback delay only.

In this letter, a simplified algorithm for designing the precoder at the transmitter for multiple antenna systems with receive correlation over Ricean fading channels is proposed. From the common property of the asymptotic solutions of high and low SNR's, a simple power allocation structure is proposed, which greatly reduce the computational complexity in comparison with the existing precoder schemes. For the case of two transmit antennas, it becomes an optimal precoder. A simple enhancement procedure, which refines the power allocation to further improve the symbol error rate (SER) performance, is also derived. Unlike most of the existing precoder schemes which need expensive computational complexity but cannot guarantee the iterative algorithms to converge, the proposed algorithm is computationally efficient and can converge to the solution fast. Simulation results show that the proposed scheme achieves SER performance close to that of the optimal one, but requires less computational complexity with fast convergence.

We propose a novel adaptive transmission scheme for space-time coded multiple-input multiple-output beamforming systems with imperfect channel state information at the transmitter, of which the signal constellation, total transmit power (temporal power), and power allocation among eigen-beams (spatial power) are jointly adapted to maximize the average spectral efficiency, subject to a target bit-error-rate and an average power constraint. The power allocation over the spatial and temporal domains makes the traditional approach of partitioning the received signal-to-noise ratio (SNR) inapplicable to the above design problem. By introducing a new variable, called as effective signal-to-noise-to-modulation ratio (ESNMR), we derive a rate-selection policy by partitioning the range of the ESNMR with an optimal set of thresholds. A closed-form temporal power control policy and a simple spatial power allocation algorithm are also obtained. Numerical results demonstrate that the new adaptive transmission scheme yields a significant performance gain over existing adaptation systems.

This paper presents a strategy to allocate power both in space and time for beamforming orthogonal space-time block coded multi antenna systems with imperfect channel state information at the transmitter. Most of the existing schemes focus on spatial-only power allocation, while the temporal dimension is not well exploited. The proposed space-time strategy is developed by decoupling the temporal and spatial power calculation through asymptotic analysis, resulting in efficient computation. Numerical results demonstrate that the proposed scheme has considerable performance gains over the existing power adaptation schemes.

The performance of multi-input multi-output (MIMO) systems with variable-power (VP) adaptive modulation (AM) and antenna selection (AS) over Rayleigh fading channels for both perfect and imperfect channel state information (CSI) is presented. In this study, the optimum fading gain switching thresholds for attaining maximum spectrum efficiency (SE) subject to a target bit error rate (BER) and an average power constraint are derived. It is shown that the Lagrange multiplier in the constrained SE optimisation is unique if the existence conditions for MIMO systems with transmit AS and receive maximal-ratio combining under perfect or imperfect CSI are satisfied. By using the switching thresholds, we obtain closed-form expressions of the SE and average BER for VP and constant-power (CP) systems. Simulation results for SE and BER are in good agreement with the theoretical analysis. The results show that the VP-AM system using AS with the derived VP control scheme provides better SE than the CP counterpart and the VP-AM system with space-time coding and can fulfil the target BER for different signal-to-noise ratios (SNRs).

This paper presents a closed-form power allocation scheme for space-time coded multiple-antenna systems with imperfect channel state information at the transmitter. The proposed scheme is based on a so-called compressed signal-to-noise ratio (CSNR) criterion, where a single compression factor is used to minimize the bit-error-rate (BER). The obtained closed-form solution is computational efficient and achieves almost the same performance as the existing optimal approach which requires numerical search, and outperforms the existing suboptimal closed-form algorithm.

This paper presents a suboptimal modified code tracking loop (SO-MCTL) that mitigates the effect of multiple- access interference (MAI) for synchronous direct-sequence code division multiple access (DS-CDMA) systems. Weighted linear combinations of the interfering users' signals are introduced into the local signals to facilitate the optimization of the tracking performance. Instead of search for the optimal weighting vectors that yields the minimum mean square tracking error, the mean square tracking error is minimized with respect to the weighting parameters given the constraints that ensure a stable bias-free S-curve. Simulations based on Gold codes compare the S-curve and root-mean-square (rms) tracking error of the SO-MCTL and traditional modified code tracking loop (T-MCTL). It is shown that SO-MCTL outperforms T-MCTL greatly in that it provides unbiased code tracking with reduced mean square tracking error and near-far resistance.

The high peak-to-average power ratio (PAR) encountered in OFDM system has been a major obstacle in the implementation of power efficient transmitter. In this paper, we present a new active constellation extension (ACE) based convex optimization algorithm which reduces PAR through convex programming. In comparison with previous convex programming method, our method greatly reduces the complexity and keeps the bit-error-rate (BER) performance. Moreover, our method can be combined with other clipping based ACE algorithms to further reduce the complexity with a slight performance degradation. Simulation results are given in this paper which show our method outperforms other ACE-based algorithms.

It is well known that visual cues of lip movement contain important speech relevant information. This paper presents an automatic lipreading system for small vocabulary speech recognition tasks. Using the lip segmentation and modeling techniques we developed earlier, we obtain a visual feature vector composed of outer and inner mouth features from the lip image sequence for recognition. A spline representation is employed to transform the discrete-time sampled features from the video frames into the continuous domain. The spline coefficients in the same word class are constrained to have similar expression and are estimated from the training data by the EM algorithm. For the multiple-speaker/speaker-independent recognition task, an adaptive multimodel approach is proposed to handle the variations caused by various talking styles. After building the appropriate word models from the spline coefficients, a maximum likelihood classification approach is taken for the recognition. Lip image sequences of English digits from 0 to 9 have been collected for the recognition test. Two widely used classification methods, HMM and RDA, have been adopted for comparison and the results demonstrate that the proposed algorithm deliver the best performance among these methods.

In this paper, we present an iterative frequency domain equalization method for orthogonal frequency division multiplexing (OFDM) over doubly selective channels. We approximate the channel by the oversampling basis expansion model (OBEM) because it can increase the model accuracy for the systems with high mobility However, the OBEM will create more serious inter-carrier interference (ICI) than the nonoversampling BEM. We propose a new iterative method to suppress the ICI effects in the frequency domain. The channel frequency response matrix (CFRM) is divided into two parts: a main band matrix as target CFRM and an out-of-band matrix as interference CFRM. At each iteration, the interference signals caused by interference CFRM are cancelled from the received signal and the target CFRM deduces the equalizer to obtain the signal estimate. Simulation results show that our proposed method outperforms the linear MMSE equalizer and has lower complexity.

In this paper we present a modified de-correlated delay lock loop (MD-DLL) for synchronous direct-sequence code division multiple access (DS-CDMA) systems. With appropriate choice of weighting and offset parameters of the local signals, the proposed MD-DLL scheme solves the tracking bias problems in de-correlated delay lock loop (D-DLL) under low SNR condition. By diminishing multiple access interference (MAI) at the on-time code position instant and minimizing the weighting imbalance between the two local signals, it is shown that MD-DLL achieves a better S-curve with negligible tracking bias and good near far resistance compared with the traditional delay lock loop (T-DLL) and D-DLL. Computer simulations based on Gold code sequences are used to verify the findings.

This paper presents a novel preamble-based synchronization scheme for orthogonal frequency division multiplexing systems over time-variant multipath fading channels. A new timing metric is derived with the use of a local synchronizing sequence (LSS). We propose one specifically designed training sequence which consists of two segments of equal length where each segment is constructed from a different pseudo-noise (PN) sequence. The advantages of using the LSS are twofold: multipath effects are reduced in the new timing metric and the metric trajectory is impulse-like. The new algorithm is shown to provide excellent performance in timing estimation even in severe time-variant fading channels. As for frequency synchronization, a two-step approach handles both fractional and integer frequency offset providing a large frequency acquisition range without loss of estimation accuracy.

This paper proposes a novel model compensation method for a robust feature extraction technique based on SNR-dependent Nonuniform Spectral Compression (SNSC). The SNSC method is a spectral transformation which resembles human's intensity-to-loudness conversion and de-emphasizes the contributions from noisy spectral components to features. In this paper, we propose a new compressed mismatch function which models the effect of the noise onto the clean speech in the Log-spectral domain together with the SNSC. Based on this mismatch function, a new model compensation procedure is derived. The procedure needs a compensated model of no compression to start with. It is shown that the new model compensation using the Vector Taylor Series method (VTS) for the compensated uncompressed model, remarkable recognition performances at low signal-to-noise ratio (SNR) can be obtained for different additive noises at the expense of slight increase in the computational complexity in comparison with the VTS.

This paper presents a new receive scheme applied to orthogonal frequency-division multiplexing (OFDM) systems for reducing inter-carrier interference (ICI) caused by frequency offset at the receiver. The proposed scheme uses a class of time domain second order polynomial Nyquist window on the receive side. The signal to interference plus noise ratio (SINR) and bit error ratio (BER) performances of the receiver are analyzed in this paper. We also propose a method to find the optimum window parameters to maximize the SINR of the receiver thus providing an adaptive receiving capability. The results show that the proposed method that enables the use of unity roll-off factor provides better SINR and BER than conventional receive methods whose roll-off factors are generally less than one. The new second order polynomial window is shown to provide better performance than raised cosine and "better than" Nyquist window.

In this paper, we further investigate the convergence properties of the generalized back-propagation algorithm using magnified gradient function (MGFPROP). The idea of MGFPROP is to increase the convergence rate by magnifying the gradient function of the activation function. The algorithm can effectively speed up the convergence rate and reduce the chance of being trapped in premature saturation. From the convergence analysis, it is shown that MGFPROP retains the gradient descent property, gives faster convergence and has better global searching capability than that of the back-propagation algorithm.

A generalized semi-blind channel estimator for OFDM using PSK signaling was proposed by the authors. It gives a significant performance gain over pilot-aided channel estimators and other semi-blind method. However, this algorithm does not work satisfactorily for the signaling of non-constant modulus. In this paper, we modify the algorithm to make it applicable to TCM-OFDM systems with biased-PSK-subset partitioning. The TCM-OFDM with 16-QAM is used for performance evaluation. Simulation results show that the generalized semi-blind channel estimation gives a significant gain over pilot-aided channel estimator and its PSK counterpart.

Parallel model combination (PMC) techniques have been successfully and popularly used in noisy speech recognition. Previous PMC methods such as the log-normal PMC and the log-add PMC generally use approximate compensation for delta parameters. In this paper we propose a PMC scheme with improved delta parameter compensation (PMC-IDC), which employs the log-normal PMC method for the static distribution compensation and a new delta parameter compensation. By using this new PMC-IDC scheme, improvement over the other PMC methods can be obtained in different additive noises at the expense of slight increase in computational complexity.

This paper presents a mean square error (MSE) analysis of the recursive least square (RLS) algorithm for system identification over wide-sense stationary uncorrelated scattering (WSSUS) fading channels. A new sum-of-sinusoids (SOS) model for modeling fading channel is proposed. Unlike most of RLS analyses, our analysis takes the correlation of the inverse of correlation matrix into account and hence yields an improved recursive formula for the MSE, which can be computed recursively with the information of the second order statistics of fading channels. The steady state MSE of RLS is derived for Clarke's model. It is shown that the MSE analysis using the SOS model have much better agreement with experimental results than the AR model.

In this paper, two modifications on the conventional back-propagation algorithm for feedforward multi-layer neural networks are presented. One modification is based on the calculation of the gradient function, while the other one is the use of evolution strategies on connection weights into the gradient search algorithm. From simulation results, the new modified algorithm always converges to the global optimal solution with better performance when compared with other fast learning algorithms and global search methods.

It is known that the perceived loudness of a tone signal by a human is spectrally masked by background noise. This masking effect causes not only a shift of just-audible sound pressure level of the tone, but also produces a masked loudness function having steeper slope than the unmasked one. This masking property of perceived loudness stimulates us to propose a new mel-scale-based feature extraction method with non-uniform spectral compression for speech recognition in noisy environments. In this method, the speech power spectrum is to undergo mel-scaled band-pass filtering, as in the standard MFCC front-end. However, the energies of the outputs of the filters are compressed by different root values defined by a compression function. This compression function is a function of the SNR in each filter band. Using this new scheme of SNR-dependent non-uniform spectral compression (SNSC) for mel-scaled filter-bank-based cepstral coefficients, substantial improvement is found for recognition in different noisy environments, as compared to the standard MFCC and features derived with cubic root spectral compression.

Authentication system solely based on visual lip information is of advantage since the uttering characteristics/manner is unique to individual. This paper presents the results of studying for the most appropriate features extracted from the lip motion for person authentication application. The geometric-based, shape-based and inner lip features are derived from a 14-point active shape model (ASM) lip model. The dynamic features reflecting the differential changes of the feature parameters are also considered in the experiment. A database consists of 40 speakers and each of their utterance last for 3 seconds. Various combinations of features obtained by analyzing this database are fed into a hidden Markov model (HMM) classifier for processing. It is observed that the best result has been obtained when both shape-based features and inner lip features are used.

Lip image analysis has attracted much interest in recent years because some important speech information is contained in the shape and movement of the lip. To extract such information from the images, accurate and robust lip region segmentation is of vital importance. However, most of the current lip segmentation methods fail to provide accurate results if the person has a beard. We propose a "one object, multiple background" clustering method to solve the problem. Since the non-lip region becomes inhomogeneous in the presence of a beard, multiple background clusters can produce better fitting to a rather complex background region than a single cluster. Spatial information in terms of the physical distance towards the lip center is incorporated to enhance the differentiation between the lip and background region. Experimental results demonstrate that our algorithm provides accurate lip segmentation results for images with beards.

This paper presents a new approach using deterministic weight modification (DWM) to speed up the convergence rate effectively and improve the global convergence capability of the standard and modified back-propagation (BP) algorithms. The main idea of DWM is to reduce the system error by changing the weights of a multi-layered feed-forward neural network in a deterministic way. Simulation results show that the performance of DWM is better than BP and other modified BP algorithms for a number of learning problems.

Total least square linear prediction has been successfully applied to frequency estimation for closely spaced sinusoids. In low signal to noise ratio, the resolving ability of TLS is degraded and extraneous roots of the predictor are close to unit circle. Hence the performance of total least square is severely degraded in low SNR. In this paper, a generalized total least squares method with a new root selection criterion, which is based on the envelope of the signal spectrum, is presented. An optimum procedure is introduced to provide a TLS solution that can perform closer to Cramer-Rao bound, particularly in low SNR.

Recently, lip image analysis has received much attention because its visual information is shown to provide improvement for speech recognition and speaker authentication. Lip image segmentation plays an important role in lip image analysis. In this paper, a new fuzzy clustering method for lip image segmentation is presented. This clustering method takes both the color information and the spatial distance into account while most of the current clustering methods only deal with the former. In this method, a new dissimilarity measure, which integrates the color dissimilarity and the spatial distance in terms of an elliptic shape function, is introduced. Because of the presence of the elliptic shape function, the new measure is able to differentiate the pixels having similar color information but located in different regions. A new iterative algorithm for the determination of the membership and centroid for each class is derived, which is shown to provide good differentiation between the lip region and the nonlip region. Experimental results show that the new algorithm yields better membership distribution and lip shape than the standard fuzzy c-means algorithm and four other methods investigated in the paper.

This paper presents two novel approaches, backpropagation (BP) with magnified gradient function (MGFPROP) and deterministic weight modification (DWM), to speed up the convergence rate and improve the global convergence capability of the standard BP learning algorithm. The purpose of MGFPROP is to increase the convergence rate by magnifying the gradient function of the activation function, while the main objective of DWM is to reduce the system error by changing the weights of a multilayered feedforward neural network in a deterministic way. Simulation results show that the performance of the above two approaches is better than BP and other modified BP algorithms for a number of learning problems. Moreover, the integration of the above two approaches forming a new algorithm called MDPROP, can further improve the performance of MGFPROP and DWM. From our simulation results, the MDPROP algorithm always outperforms BP and other modified BP algorithms in terms of convergence rate and global convergence capability.

Loudness is a function of sound pressure level. The power law used in approximating the loudness function has an exponent that depends on the bandwidth of the sound signal. This exponent decreases from about 0.3 for a narrow band tone to 0.23 for a broadband uniform-exciting noise. Exploiting this property of psychoacoustics of hearing, this paper proposes a new feature extraction method for robust speech recognition for FFT-based methods. In the method, larger energy compression is applied to broadband-like high frequency bands of the power spectrum of each frame, instead of a fixed compression for all frequency bands as in root cepstral analysis or perceptually based linear prediction (PLP). Further to this, those sound segments or frames having broadband characteristics like those of fricatives are given larger compression as well. The frame energy is used as the index to determine the degree of compression. By using this new scheme of non-uniform spectral compression, significant improvement in recognition accuracy is obtained, especially in very low SNR, under white noise environment.

A new lip contour extraction algorithm that combines the merits of the point-based lip model and the parametric lip model is presented. A 16-point lip model is used to describe the lip contour. With the aid of the FCMS (fuzzy clustering method incorporating shape function), a robust probability map is generated and a region-based cost function can be established. An iterative optimization procedure has been developed to fit the lip model to the probability map. In each iteration, the adjustment of the 16 lip points is governed by three pieces of quadratic curves which constrain the points to form a physical lip shape. Experimental results show that the proposed approach provides satisfactory results for 5,000 lip images of over 20 individuals. A real-time lip contour extraction system has also been implemented.

This paper presents a modified back-propagation algorithm using magnified gradient function (MGFPROP), which can effectively speed up the convergence rate and improve the global convergence capability of back-propagation. The purpose of MGFPROP is to increase the convergence rate by magnifying the gradient function of the activation function. From the convergence analysis, it is shown that the new algorithm retains the gradient descent property but gives faster convergence than that of the back-propagation algorithm. Simulation results show that, in terms of the convergence rate and the percentage of global convergence, the new algorithm always outperforms the standard back-propagation algorithm and other competing techniques.

A gradient based variable forgetting factor (GVFF) RLS lattice (RLSL) algorithm is introduced in this paper. The steepest descent approach is used to control the forgetting factor which is based on the dynamic equation of the gradient of the mean square error. Compared with the standard RLSL algorithm, GVFF-RLSL algorithm gives fast tracking with a small mean square model error and its performance is not degraded much even in low signal-to-noise ratios (SNR) for time varying systems.

In this paper, we describe the application of a novel spatial fuzzy clustering algorithm to the lip segmentation problem. The proposed spatial fuzzy clustering algorithm is able to take into account both the distributions of data in feature space and the spatial interactions between neighboring pixels during clustering. By appropriate pre- and postprocessing utilizing the color and shape properties of the lip region, successful segmentation of most lip images is possible. Comparative study with some existing lip segmentation algorithms such as the hue filtering algorithm and the fuzzy entropy histogram thresholding algorithm has demonstrated the superior performance of our method.

In environment with impulsive noise, most learning algorithms are encountered difficulty in distinguishing the nature of large error signal, whether caused by the impulse noise or large model error. Consequently, they suffer from slow convergence or large misadjustment. A new gradient based variable forgetting factor nonlinear RLS algorithm uses correlation function of error signal with nonzero lags (GCVFF) is introduced. The correlation of nonzero lags maintains the sensitivity of the algorithm responding to the model error and becomes sluggish to the impulse noise. Simulation results show that it achieves fast convergence speed and small misadjustment and outperforms other variable forgetting factor (VFF) RLS algorithms.

In lip segmentation, most of the clustering methods are based on color information. In this paper, a new fuzzy clustering method that combines the color information and the shape information in the measure is proposed. An elliptic shape function is introduced in the measure that modifies the measure according to the distance of the pixel from the lip center with which the differentiation between the lip and non-lip regions is enhanced. Compared with the standard fuzzy c-means algorithm, experimental results show that the new algorithm has better membership distribution and shape. This method is applicable to other images with a well-predicted shape.

In this paper we introduce a design of a phase-modulation spreading sequence for a spread spectrum multiple access communication system with full response continuous phase modulation. The continuous phase modulation with the phase-modulation spreading sequences in the CDMA communication system is shown to provide a desirable spectral property and lower bit-error-rate in comparison with BPSK over Rayleigh fading channel.

This paper continues the discussion of weight evolution algorithm for solving the local minimum problem of back-propagation by changing the weights of a multi-layer neural network in a deterministic way. During the learning phase of back-propagation, the network weights are adjusted intentionally in order to have an improvement in system performance. The idea is to work backward from the error components and the system outputs to deduce a deterministic perturbation on particular network weights for optimization purpose. Simulation results show that the weight evolution algorithm always outperforms the other traditional methods in achieving global convergence. From mathematical analysis, it can be found that the weight evolution between the hidden and output layers can accelerate the convergence speed, whereas the weight evolution between the input and hidden layers can assist in solving the local minima problem.

In an impulsive noise environment, most learning algorithms encounter difficulty in distinguishing the nature of a large error signal, whether caused by the impulse noise or model error. Consequently, they suffer from large misadjustment or otherwise slow convergence. A new nonlinear RLS (VFF-NRLS) adaptive algorithm with variable forgetting factor for FIR filters is introduced. In this algorithm, the autocorrelations of non-zero lags, which is insensitive to white noise, is used to control the forgetting factor of the nonlinear RLS. This scheme makes the algorithm have fast tracking capability and small misadjustment. By experimental results, it is shown that the new algorithm can outperform other RLS algorithms.

This paper presents a new optimization procedure for extracting the point-based lip contour using the active shape model (ASM). A 14-point ASM lip model is used to describe the lip contour. With the aid of fuzzy clustering analysis, a probability map of the color lip image is obtained and a region-based cost function is established. The new optimization procedure operates on the spatial domain (actual contour points) and all the points are pulled towards their desirable locations in each iteration. Hence, the lip contour evolution becomes better controlled and consequently fast convergence is achieved. The new procedure can also achieve real-time performance on lip contour extraction and tracking from lip image sequence.

This paper introduces a new weight evolution algorithm to find the global minimum of the error function in a multi-layered neural network. During the learning phase of backpropagation, the network weights are adjusted intentionally in order to have an improvement in system performance. By looking at the system outputs of the nodes, it is possible to adjust some of the network weights deterministically so as to achieve an overall reduction in system error. The idea is to work backward from the error components and the system outputs to deduce a deterministic perturbation on particular network weights for optimization purposes. Using the new algorithm, it is found that the weight evolution between the hidden and output layer can accelerate the convergence speed, whereas the weight evolution between the input layer and the hidden layer can assist in solving the local minima problem.

The use of visual information from lip movements can improve the accuracy and robustness of a speech recognition system. Accurate extraction of visual features associated with the lips is thus very important. In this paper, we present a deformable model-based technique for lip contour extraction from color lip images. The geometric lip model we propose is able to fit different lip shapes. Our method starts by generating a probability map of the lip image using spatial fuzzy clustering. Then the optimum set of model parameters that partitions a lip probability map into lip region and non-lip region is found, such that the joint probability of the two regions is maximised. Experimental results indicate that accurate and robust lip contour extraction is possible using this approach.

Introduces a learning algorithm for multi-layered feedforward network-weight evolution algorithm with deterministic perturbation. During the learning phase of the gradient algorithm (such as backpropagation), the network weights are adjusted intentionally in order to have an improvement in system performance. The intention is to reduce the overall system error after every weight update. By looking at the error component, it is possible to adjust some of the network weights dramatically so as to have an overall reduction in system error. Using the deterministic perturbation, it is found that the weight evolution between the hidden and output layer can accelerate the convergence speed, whereas the weight evolution between the input layer and the hidden layer can assist in solving the local minima problem.

A new variable step size scheme for the least-mean-square (LMS) algorithm is proposed in this paper. The idea is basically a kind of evolutionary strategies. The step size candidates are generated and evaluated by calculating a square error measure based on a priori and a posteriori errors. The fittest candidate is selected for subsequent adaptation. The composition of the square error measure is regulated according to the mean square error so as to provide fast converging and tracking capability. The convergence performance is significantly improved and is less sensitive to eigenvalue spread.

Concerns feedforward neural net training. The conventional backpropagation algorithm will always get stuck into local minima and converge very slowly. Other fast algorithms can increase the convergence speed however they still converge to local minima. We introduce a new hybrid algorithm with the use of weight evolution into the generalized back-propagation method. The hybrid algorithm further improve the convergence rate and the global convergence capability, it ensures the convergence to a global minimum in a compact region of a weight vector space.

A clustering-based lip segmentation algorithm is described here. The fuzzy clustering algorithm presented here is able to take into account the local smoothness property of image data. The objective functional of our algorithm utilizes a new distance metric that takes into account the influence of the neighboring pixels on the centre pixel in a 3 by 3 window. Computational steps involved in the segmentation of color lip image and segmentation result for a set of color lip images are given.

The general formula for evaluating the bit-error probability of M-ary differential phase-shift keying with L diversity equal-gain combining over frequency nonselective m-distributed, uncorrelated Nakagami fading channels and correlated Gaussian noise is presented. The evaluation usually takes a one-dimensional integration and this integration can be reduced to a closed form for integer m by applying the residue theorem.

This paper presents a frequency weighted generalized total least squares linear prediction for estimating closely spaced sinusoids. In this method, the received data is first processed by a pole-zero prefilter and then a generalized total least squares linear prediction is applied to the prefiltered signal. A procedure of optimizing the generalized solution is introduced. By computer simulations, it is shown that the solution can outperform the existing well known total least squares solutions especially in low signal-to-noise ratio.

This paper presents a general total least squares (GTLS) solution for linear prediction to estimate closely spaced sinusoids. It is found that the TLS prediction error is not a good criterion to provide a robust solution. In this paper, a frequency weighted prediction error approach is introduced. Experimental results show that the GTLS solution based on the frequency weighted prediction error can give a robust performance even in very low SNR.

Vector quantization (VQ) is usually adopted to quantize parameters of multiple dimensions. It uses certain meaningful vector measures for codevector selection and codebook training to achieve bit reduction, at the expense of, however, higher computation and memory requirement. Due to the aforementioned shortcomings, in practical applications, scalar quantization (SQ) is still applied to quantize each of the multidimensional parameters individually. In this paper, we propose to apply a vector measure method to the table lookup in scalar quantization so as to improve its coding efficiency without increasing storage requirement. Two LSF scalar quantizers are investigated in which the weighted Euclidean distance is used as a vector measure to improve the performance of both quantizers.

A new variable step-size algorithm which has a framework similar to the variable step (VS) adaptive filter algorithm given by R. W. Harris et al. (1986) is introduced. The methodology is to periodically invoke a generate-and-evaluate function during the adaptation. The new function will first generate a new step-size by increasing or decreasing the current step-size value by a factor of /spl alpha/ according to the sign change of the gradient. Then the performance of the current step-size and the newly generated step-size will be evaluated and the better one will be chosen for the next stage of adaptation. Empirically, the value of /spl alpha/ is almost linearly related to the invocation period T. The smaller values of /spl alpha/ and T give better tracking capability. Simulation results show that the new algorithm outperforms the LMS with fixed step-size and the VS algorithm in terms of convergence and steady-state misadjustment. In addition, the new algorithm can be implemented with only modest increase in complexity over the LMS algorithm by using a fast implementation scheme.

The classical learning algorithms for adaptive IIR filtering, such as gradient-descent algorithm and least square techniques, suffer from several weaknesses. First, the convergence time is too long even for low order filters. Second, the algorithms fail to converge to the global optimum when the error function is multimodal. To tackle the above difficulties, a new learning algorithm for adaptive IIR filtering is proposed. In this paper, the genetic search is introduced into the gradient-descent algorithm, such as the least-mean-square (LMS) algorithm, so as to provide global search capability and to further improve its convergence speed. In addition, the new algorithm is also applied to lattice structure of IIR filters for providing a more stable behavior.>

This paper proposes a new variable step-size algorithm using an evolutionary-type searching approach. The step-sizes of the adaptive filter are evolved regularly in a controlled manner to reduce the squared estimation error at each iteration. The algorithm improves the rate of convergence and achieves a low steady-state misadjustment. The performance analysis of the adaptive filter and simulation results are presented in the paper. The results show that the new algorithm outperforms the least mean square (LMS) algorithm with constant step-size considerably.>

This paper investigates the channel equalization by using a recursive lattice filter cascaded with a sigmoidal activation function. A new evolutionary search algorithm is proposed for the adaptation of the equalizer coefficients. The new algorithm is based on random evolution of the filter coefficients during the adaptation phase of the gradient lattice algorithm. The proposed method can greatly improve the convergence speed of the gradient method and provide global search capability so as to give lower error performance. In addition, the new algorithm maintains the essence and simplicity of the original gradient descent method. The equalizer together with the new algorithm is shown to have lower bit error rate.>

A fractionally spaced equalizer structure is presented. This structure is a cascade form that comprises a predictor and a phase-splitting fractionally spaced equalizer (PS-FSE). The function of the predictor is to preequalize the spectrum of the channel output and reduce the eigenvalue spread of the correlation matrix of the data so as to improve the convergence of the PS-FSE. Comparing the new structure to the complex fractionally spaced equalizer (CFSE), the computations are significantly reduced and the convergence remains almost the same. A new sign algorithm which employs a three-level quantizer is introduced to update the filter coefficients for further simplifying the computations. An optimum three-level quantizer is developed through the detailed analysis of the algorithm. The performance of this three-level sign algorithm (TLSA) is almost the same as that of the LMS algorithm. Simulations prove that the new structure together with the TLSA performs satisfactorily and is a good structure for passband equalization.>

A single channel spectrum estimation is transformed to a multichannel problem. If λ is the pole of the autoregressive model and L is the dimension of the channels, then λLis the resulting pole location obtained from the multichannel approach. The simulations illustrate that the new method is superior over the Burg's method for the lowpass signals, regarding to the resolution.