Discriminative temporal feature extraction for robust speechrecognition

A discriminative temporal feature processing method for robust speech recognition is presented by combining the knowledge and the statistical methods. The cepstral features are first filtered by a RASTA method based on human hearing perception and then processed using the minimum classification error algorithm. Improved recognition performance can be achieved in both quiet and noisy environments     

Reduced complexity implementation of Bayesian equaliser using localRBF network for channel equalisation problem

The authors examine a method for reducing the implementation complexity of the RBF Bayesian equaliser using model selection. The selection process is based on finding a subset model to approximate the response of the full RBF model for the current input vector, and not for the entire input space. Using such a scheme, for cases in which the channel equalisation problem is non-stationary, the requirement for updating all the centre locations is removed, and the implementation complexity is reduced. Using computer simulations, we show that the number of centres can be greatly reduced without compromising classification performance     

Fixed-lag smoother for digital channel equalisation

A new equaliser for digital communication channels, which carries out a discrete-time fixed-lag smoothing of the state vector of the channel, modelled as a tapped delay line of finite length, is proposed. The smoothed estimate uses and improves the filtered estimate of the conventional Kalman-filter equaliser.     

Variation of features of interframe dependent HMM for speechrecognition

The effects are explored of using different dynamic features in conjunction with an HMM that permits a dependency on both preceeding and succeeding frames. In particular, features which capture dynamic information are varied in the interframe dependent HMM. A recognition accuracy of 93.9% for the Connex speaker-independent E-set was obtained using tied states     

A Bayesian classification approach with application to speechrecognition

A Bayesian approach to classification of parametric information sources whose statistics are not explicitly given is studied and applied to recognition of speech signals based upon Markov modeling. A classifier based on generalized likelihood ratios, which depends only on the available training and testing data, is developed and shown to be optimal in the sense of achieving the highest asymptotic exponential rate of decay of the error probability. The proposed approach is compared to the standard classification approach used in speech recognition, in which the parameters for the sources are first estimated from the given training data, and then the maximum a posteriori decision rule is applied using the estimated statistics     

Joint channel estimation and turbo equalisation for MIMO-OFDM-IM systems

Multiple-Input Multiple-Output (MIMO) communications are frequently employed to improve the transmitted data rate and the link quality. Index modulated orthogonal frequency division multiplexing (OFDM-IM) improves the error rate performance and the peak-to-average power ratio (PAPR) compared with those of the conventional OFDM system due to the activation of partial subcarriers. The MIMO OFDM-IM can transmit additional information bits via the indices of active subcarriers. Also, in order to reduce the transmission power of the OFDM system, the MIMO OFDM-IM scheme can be employed to approach the demanded data transmission rate and the error rate performance. Multiple-input multiple-output orthogonal frequency division multiplexing index modulation (MIMO-OFDM-IM) is an effective multicarrier transmission scheme and can be proposed as an alternative to conventional MIMO-OFDM system. In this scheme, OFDM-IM is combined with MIMO transmission to take the benefits of these two techniques. In this paper, we propose a joint channel estimation and turbo equalisation receiver for MIMO-OFDM-IM system. Some simulation examples are given to demonstrate the effectiveness of the proposed receiver.     

Sequential algorithm for channel equalisation employing a stack

A sequential algorithm based on a stack, and applicable to the problem of recovering digital data after degradation by intersymbol interference and additive noise, is described. Results are presented showing a tradeoff between performance and equipment complexity, with performance asymptotic to the maximum-likelihood-sequence (Viterbi) estimate when the storage is large.     

Active feed-forward channel equalisation for chirped pulsewavelength division multiplexing

A channel equalisation scheme for chirped-pulse wavelength-division-multiplexing (CPWDM) is described and implemented. The equalisation is realised using a simple feed-forward circuit that dynamically varies the loss of the modulator, which is used to sequentially encode channels in CPWDM to actively adjust the channel strengths. A 50% increase in the number of channels within 3 dB power variation is achieved for our laser spectrum     

Bayesian blind and semi-blind equalisation of channels with Markovinputs

An original full Bayesian approach is developed for blind and semi-blind equalisation of fading channels with Markov inputs. The sequence of discrete symbols is estimated according to a marginal maximum a posteriori criterion; the other unknown parameters are regarded as random nuisance parameters and are integrated out analytically. A batch algorithm is proposed to maximise the marginal posterior distribution. Simulation results are presented to demonstrate the effectiveness of the method     

Block-based fuzzy step size nlms algorithms for subband adaptive channel equalisation

Judicious selection of the step size parameter is crucial for adaptive algorithms to strike a good balance between convergence speed and misadjustment. The fuzzy step size (FSS) technique has been shown to improve the performance of the classical fixed step size and variable step size (VSS) normalised least mean square (NLMS) algorithms. The performance of the FSS technique in the context of subband adaptive equalisation is analysed and two novel block-based fuzzy step size (BFSS) strategies for the NLMS algorithm, namely fixed block fuzzy step size (FBFSS) and adaptive block fuzzy step size (ABFSS) are proposed. By exploiting the nature of gradient noise inherent in stochastic gradient algorithms, these strategies are shown to substantially reduce the computational complexity of the conventional FSS technique without sacrificing the convergence speed and steady-state performance. Instead of updating the step size at every iteration, the proposed techniques adjust the step size based on the instantaneous squared error once over a block length. Design methodology and guidelines that lead to good performance for the algorithms are given.     

Adaptive retraining equalisation based on channel change detections

By adaptively detecting abrupt changes in the channel tap coefficients and requesting the training sequence to be transmitted whenever changes are detected, an adaptive retraining equaliser has been designed. The performance of the equaliser is evaluated by numerical simulations, and the results show that the equaliser outperforms the traditional periodical retraining equaliser and requires fewer training sequences.     

Fast method of channel equalisation for speech signals and itsimplementation on a DSP

Blind equalisation of a speech signal that has been passed over a linear filter can be achieved by estimating the poles of the signal and separating the stationary poles due to the filter from the time varying poles due to the speech. However, identification of the position of the stationary poles, conventionally done by pole clustering, is unreliable and slow. A new algorithm for the identification of stationary poles is presented which is more accurate and faster than clustering     

2-D adaptive CPWL filter for 2-D nonlinear channel equalisation and image restoration

A 2-D adaptive piecewise-linear equaliser is proposed. As an alternative of the 2-D adaptive Volterra equaliser, the proposed equaliser has advantages in its suitability for cases of strong nonlinearity and saving implementation and computation cost. An experiment examples is presented which demonstrates the superior behaviour of this method over either a linear method or the Volterra method in inverse modelling an unknown 2-D channel with blurring and a common point-wise nonlinearity and restoring images degraded by this channel.<>     

Online blind equalisation algorithm of an FIR MIMO channel system for non-stationary signals

Blind equalisation of an FIR multi-input multi-output channel system is an important task for numerous applications such as speech separation, de-reverberation, communication, signal processing and control, etc. In this paper, a cost function with the knowledge of correlation is reconstructed and a new online algorithm derived with a natural gradient search method for blind source separation of convolutional mixtures. Its implementation is simple and practical. Furthermore, the equivariance property is possessed by the algorithm. Simulations indicate the ability of the algorithm to perform blind equalisation under the weaker condition (the FIR system is equalisable) and also to make speech separation and de-reverberation simultaneous.     

Bayesian techniques for known channel deconvolution

This paper introduces a new family of deconvolution filters for digital communications subject to severe intersymbol interference. These fixed lag smoothing filters for known channel demodulation are called Bayesian filters. Bayesian filters are derived using a new approach to suboptimal recursive minimum mean square error estimation for non-Gaussian processes. The family of Bayesian filters interpolates between the optimum fixed lag linear filter (i.e., the Kalman filter) and the optimum fixed lag symbol-by-symbol demodulator in both performance and complexity. The complexity of the Bayesian filter is exponential in a parameter, typically chosen smaller than the channel length and the filter lag. Hence, the Bayesian filter decouples the channel length and the filter lag from the exponential complexity in these parameters found in many other high performance demodulation algorithms. Simulations characterize the performance and compare the Bayesian filter to both optimal and reduced complexity demodulation algorithms     

Cepstrum third-order normalisation method for noisy speechrecognition

A new cepstrum normalisation method is proposed which can be used to compensate for distortion caused by additive noise. Conventional methods only compensate for the deviation of the cepstral mean and/or variance. However, deviations of higher order moments also exist in noisy speech signals. The proposed method normalises the cepstrum up to its third-order moment, providing closer probability density functions between clean and noisy cepstra than is possible using conventional methods. From the speaker-independent isolated-word recognition experiments, it is shown that the proposed method gives improved performance compared with that of conventional methods, especially in heavy noise environments     

Bayesian inference of models and hyperparameters for robust optical-flow estimation

Selecting optimal models and hyperparameters is crucial for accurate optical-flow estimation. This paper provides a solution to the problem in a generic Bayesian framework. The method is based on a conditional model linking the image intensity function, the unknown velocity field, hyperparameters, and the prior and likelihood motion models. Inference is performed on each of the three levels of this so-defined hierarchical model by maximization of marginalized a posteriori probability distribution functions. In particular, the first level is used to achieve motion estimation in a classical a posteriori scheme. By marginalizing out the motion variable, the second level enables to infer regularization coefficients and hyperparameters of non-Gaussian M-estimators commonly used in robust statistics. The last level of the hierarchy is used for selection of the likelihood and prior motion models conditioned to the image data. The method is evaluated on image sequences of fluid flows and from the "Middlebury" database. Experiments prove that applying the proposed inference strategy yields better results than manually tuning smoothing parameters or discontinuity preserving cost functions of the state-of-the-art methods.     

Adaptive blind channel equalisation in chaotic communications by using nonlinear prediction technique

In chaotic communications, an ideal channel is often assumed. In practice, channel distortion is inevitable. In particular, in wireless chaotic communications, the channel distortion may be serious and must be compensated. An adaptive blind equalisation algorithm is proposed. The aim of the algorithm is to recover the chaotic signal transmitted through a finite impulse response (FIR) channel. The inherent characteristic of the chaotic signal, that is the high sensitivity to initial conditions, is exploited to formulate the criterion used in deriving the algorithm. The analysis of stability of the proposed algorithm is also provided     

Use of multiple vector quantisation for semicontinuous-HMM speechrecognition

Although the continuous hidden Markov model (CHMM) technique seems to be the most flexible and complete tool for speech modelling. It is not always used for the implementation of speech recognition systems because of several problems related to training and computational complexity. Thus, other simpler types of HMMs, such as discrete (DHMM) or semicontinuous (SCHMM) models, are commonly utilised with very acceptable results. Also, the superiority of continuous models over these types of HMMs is not clear. The authors' group has previously introduced the multiple vector quantisation (MVQ) technique, the main feature of which is the use of one separated VQ codebook for each recognition unit. The MVQ technique applied to DHMM models generates a new HMM modelling (basic MVQ models) that allows incorporation into the recognition dynamics of the input sequence information wasted by the discrete models in the VQ process. The authors propose a new variant of HMM models that arises from the idea of applying MVQ to SCHMM models. These are SCMVQ-HMM (semicontinuous multiple vector quantisation HMM) models that use one VQ codebook per recognition unit and several quantisation candidates for each input vector. It is shown that SCMVQ modelling is formally the closest one to CHMM, although requiring even less computation than SCHMMs. After studying several implementation issues of the MVQ technique. Such as which type of probability density function should be used, the authors show the superiority of SCMVQ models over other types of HMM models such as DHMMs, SCHMMs or the basic MVQs