The constant modulus algorithm (CMA) is an excellent technique for blind channel equalization. A signed error version of CMA (SE-CMA) and dithered signed error version (DSE-CMA) have been proposed which afford overall computational efficiency. We propose three different error functions for faster convergence. This would be essential for communication systems, which cannot afford a high startup delay or for systems, where the channel's impulse response is rapidly fluctuating. One of these algorithms relies on the idea of a variable step size, which increases the rate of convergence.

Different variants of evolutionary programming (EP) have been proposed recently to determine the order and parameters of time series models. Unlike conventional algorithms, the model order and coefficients are evolved simultaneously in evolutionary algorithms. In this paper, the performance of the different types of evolutionary algorithms was tested on standard problems. In particular, the algorithms considered in this paper were evolutionary programming, fast evolutionary programming and modified simulated evolutionary optimization, EP was used to evolve both the order and coefficients of the reduced parameter bilinear model and recurrent bilinear perceptron. These models were used for one-step prediction of four well investigated time series, namely the sunspot series, the Mackey-Glass series, the laser data series and the astrophysical data series. The performance of the algorithms was compared on the basis of the order of the model evolved and normalized mean square error. Fast evolutionary programming with recurrent bilinear perceptrons produced the best models with fewer parameters and lower normalized mean square error.

A closed-loop adaptive subband prediction architecture is presented by employing an adaptive subband filter in the prediction configuration. Some authors have suggested that applying open-loop prediction methods to subband signals can realize increased prediction gain over fullband prediction. Furthermore, the benefits of applying multirate techniques to adaptive filtering are well understood in terms of reduction of computational complexity and increased convergence speed. Thus, the closed-loop subband adaptive predictor is a novel approach that is expected to exhibit these same benefits along with the advantages of backward adaptation. Results show that the new subband predictor can produce a higher prediction gain than a similar fullband adaptive prediction filter. The proposed architecture is implemented in C++ on the Pentium processor.

This letter presents a new algorithm, fast evolutionary programming (FEP), for determining the model orders and parameters of reduced parameter bilinear (RPBL) models used for predicting nonlinear and chaotic time series. FEP is a variant of the conventional evolutionary programming (EP) algorithm with a new mutation operator. This new mutation operator enhances EP's ability to escape from local minima resulting in a significantly faster convergence to the optimal solution. Both the model order and the parameters are evolved simultaneously. Experimental results on the sunspot series and Mackey-Glass series show that FEP is capable of determining the optimal model order and, in comparison with conventional evolutionary programming, evolves models with lower normalized mean squared error.

We present a high performance neural network based multiuser detector in code division multiple access (CDMA) communications. This detector retains salient features of the annealed neural network based detector (ANNMD) and reduces its computational complexity. The BER performance of the proposed detector is close to the theoretical lower bound of the BER performance. We present a theoretical derivation of the hybrid type multiuser detector with improved hardware complexity. Extensive numerical evaluation of the proposed techniques as well as various suboptimal multiuser detectors is conducted using Monte-Carlo simulation.

The main goal of any electrocardiogram (ECG) compression algorithm is to reduce the bit rate while keeping the signal distortion at a clinically acceptable level. Percentage root mean square difference (PRD), the commonly used figure of merit, does not directly reveal whether the clinically significant ECG waveform information is preserved or not. We present the results of a study of ECG compression using an upper bound on the PRD. This bound is based on the initial performance of the algorithm and could be specified by the clinician after correlating the quality of the compressed versions of the ECG and the resulting PRD.

The article presents a new algorithm, simulated evolutionary optimization, a multi-agent stochastic search method for optimizing the coefficients of a FIR digital filter in the powers-of-two space by minimizing a sum of two cost functions, namely the squared error between discrete filter frequency response and desired frequency response and the squared error between discrete and infinite precision coefficients. The proposed algorithm requires less computation time than existing techniques based on linear programming and simulated annealing and differs from the tree search based algorithms in optimizing the coefficients simultaneously. The effectiveness of the method is demonstrated through a low-pass filter example.

A promising method is introduced for adjusting the LMS (least-mean-square) step size mu , based on the time variation of the tap weights. This method belongs to the class of diagonal step-matrix algorithms: a separate mu is used for each tap weight. Computer simulations are presented for a self-tuning filter configuration operating on a sine wave of abruptly changing frequency in Gaussian white noise. It is shown that the algorithm can converge quickly, with small error after convergence.>

An algorithm for estimating the frequencies of two sinusoids in white noise is presented. The method uses the eigenvectors of the estimated autocorrelation matrix (ACM) to yield the signal vectors. Since the signal and the eigenvectors span the same space, one can be obtained from the other through a linear transformation. The coefficients of the linear combination are derived as the eigenvectors of a 2*2 matrix whose elements are a function of the signal power, ACM dimension, and the angle between the two signal vectors. A method for estimating the unknown angle between the signal vectors is described. Knowledge of this angle completely determines the 2*2 matrix whose eigenvectors then yield the signal vectors. A periodogram can now be used to estimate the unknown frequencies from each signal vector. The method has the attractive property of separating a two-sinusoid problem into two single-sinusoid problems. Simulation results that describe the performance of this algorithm at low signal-to-noise ratios are presented.>

The problem of estimating sinusoidal frequencies at low signal-to-noise ratios (SNRs) is considered. The principal component (PC) method has a threshold SNR of about 7 dB. A lowering of this threshold can be achieved if the PC method is cascaded with the Burg algorithm. The reduction is due to the elimination of noise from the noisy data vector, by projecting it on to signal subspace. Simulation results are presented that demonstrate the improvement in performance at low SNRs.>

The problem of determining the dominant singular values in the singular value decomposition (SVD) based state-space approach to harmonic retrieval is considered. A common difficulty encountered in harmonic retrieval methods is that the covariance matrix is full rank due to noise and estimation errors, instead of the ideal low rank. Then, from the singular value decomposition of this noisy and estimated covariance matrix, a low rank approximation is normally sought by retaining the dominant singular values and zeroing out the rest. A criterion is proposed, based on the distribution of the norms of the perturbation matrix associated with the estimated covariance matrix, to identify these dominant singular values.>