Muhammad Moinuddin

Convergence and tracking analysis of a variable normalised LMF (XE-NLMF) algorithm

The least-mean-fourth (LMF) algorithm is known for its fast convergence and lower steady state error, especially in sub-Gaussian noise environments. Recent work on normalised versions of the LMF algorithm has further enhanced its stability and performance in both Gaussian and sub-Gaussian noise environments. For example, the recently developed normalised LMF (XE-NLMF) algorithm is normalised by the mixed signal and error powers, and weighted by a fixed mixed-power parameter. Unfortunately, this algorithm depends on the selection of this mixing parameter. In this work, a time-varying mixed-power parameter technique is introduced to overcome this dependency. A convergence analysis, transient analysis, and steady-state behaviour of the proposed algorithm are derived and verified through simulations. An enhancement in performance is obtained through the use of this technique in two different scenarios. Moreover, the tracking analysis of the proposed algorithm is carried out in the presence of two sources of nonstationarities: (1) carrier frequency offset between transmitter and receiver and (2) random variations in the environment. Close agreement between analysis and simulation results is obtained. The results show that, unlike in the stationary case, the steady-state excess mean-square error is not a monotonically increasing function of the step size.

Tracking analysis of the NLMS algorithm in the presence of both random and cyclic nonstationarities

Tracking analysis of the normalized least mean square (NLMS) algorithm is carried out in the presence of two sources of nonstationarities: 1) carrier frequency offset between transmitter and receiver; 2) random variations in the environment. A novel approach to this analysis is carried out using the concept of energy conservation. Close agreement between analytical analysis and simulation results is obtained. The results show that, unlike in the stationary case, the steady-state excess MSE is not a monotonically increasing function of the step size. Moreover, the ability of the adaptive algorithm to track the variations in the environment is shown to degrade with increasing frequency offset.

Automatic classification of speech and music using neural networks

The importance of automatic discrimination between speech signals and music signals has evolved as a research topic over recent years. The need to classify audio into categories such as speech or music is an important aspect of many multimedia document retrieval systems. Several approaches have been previously used to discriminate between speech and music data. In this paper, we propose the use of the mean and variance of the discrete wavelet transform in addition to other features that have been used previously for audio classification. We have used Multi-Layer Perceptron (MLP) Neural Networks as a classifier. Our initial tests have shown encouraging results that indicate the viability of our approach.

Exact Performance Analysis of the

This paper presents exact mean-square analysis of the -NLMS algorithm for circular complex correlated Gaussian input. The analysis is based on the derivation of a closed form expression for the cumulative distribution function (CDF) of random variables of the form ∥ui∥_D₁²/(ϵ + ∥ui∥_D₁²) and using that to derive the first and second moments of such variables. These moments in turn completely characterize the mean square (MS) behavior of the ϵ-NLMS in explicit closed form expressions. Both transient and steady-state behavior are analyzed. Consequently, new explicit closed-form expressions for the mean-square-error (MSE) behavior are derived. Our simulations of the transient and steady-state behavior of the filter match the expressions obtained theoretically for various degrees of input correlation and for various values of ϵ.

\epsilon

A detailed analysis of the multiple access interference (MAI) for synchronous downlink CDMA systems is carried out for BPSK signals with random signature sequences in Nakagami- fading environment with known channel phase. This analysis presents a unified approach as Nakagami- fading is a general fading distribution that includes the Rayleigh, the one-sided Gaussian, the Nakagami-, and the Rice distributions as special cases. Consequently, new explicit closed-form expressions for the probability density function (pdf ) of MAI and MAI plus noise are derived for Nakagami-, Rayleigh, one-sided Gaussian, Nakagami-, and Rician fading. Moreover, optimum coherent reception using maximum likelihood (ML) criterion is investigated based on the derived statistics of MAI plus noise and expressions for probability of bit error are obtained for these fading environments. Furthermore, a standard Gaussian approximation (SGA) is also developed for these fading environments to compare the performance of optimum receivers. Finally, extensive simulation work is carried out and shows that the theoretical predictions are very well substantiated.

-NLMS Algorithm for Colored Circular Gaussian Inputs

Since multiuser code-division multiple-access (CDMA) communications systems suffer significantly from multiple-access interference (MAI) and from classical white Gaussian noise, it is therefore necessary to consider their impact on the performance of these systems. It is well known that the learning speed of any adaptive filtering algorithm is increased by adding a constraint to it. In this paper, a constrained least-mean-square (LMS) algorithm, which incorporates the knowledge of the number of users, spreading sequence length, and additive noise variance, is developed subject to the new combined constraint comprising the MAI and noise variance for a synchronous downlink direct-sequence CDMA system. The novelty of this constraint resides in the fact that the MAI variance was never used as a constraint. In our approach, a Robbins-Monro algorithm is used to minimize the conventional mean-square-error criterion subject to the variance of the new constraint (MAI plus noise). This constrained optimization technique results in an (MAI plus noise)-constrained LMS (MNCLMS) algorithm. The MNCLMS algorithm is a type of variable step-size LMS algorithm where the step-size rule arises naturally from the constraints on MAI and noise variance. Convergence and tracking analysis of the proposed algorithm are carried out in the presence of MAI. Finally, a number of simulations are conducted to compare the performance of the MNCLMS algorithm with other adaptive algorithms.

A unified approach to BER analysis of synchronous downlink CDMA systems with random signature sequences in fading channels with known channel phase

Radial basis function neural networks are used in a variety of applications such as pattern recognition, nonlinear identification, control and time series prediction. In this paper, the learning algorithm of radial basis function neural networks is analyzed in a feedback structure. The robustness of the learning algorithm is discussed in the presence of uncertainties that might be due to noisy perturbations at the input or to modeling mismatch. An intelligent adaptation rule is developed for the learning rate of RBFNN which gives faster convergence via an estimate of error energy while giving guarantee to the l 2 stability governed by the upper bounding via small gain theorem. Simulation results are presented to support our theoretical development.

Multiple-Access Interference Plus Noise-Constrained Least Mean Square (MNCLMS) Algorithm for CDMA Systems

The learning speed of an adaptive algorithm can be improved by properly constraining the cost function of the adaptive algorithm. In this work, a noise-constrained least mean fourth (NCLMF) adaptive algorithm is proposed. The NCLMF algorithm is obtained by constraining the cost function of the standard LMF algorithm to the fourth-order moment of the additive noise. The NCLMF algorithm can be seen as a variable step-size LMF algorithm. The main aim of this work is to derive the NCLMF adaptive algorithm, analyze its convergence behavior, and assess its performance in different noise environments. Furthermore, the analysis of the proposed NCLMF algorithm is carried out using the concept of energy conservation. Finally, a number of simulation results are carried out to corroborate the theoretical findings, and as expected, improved performance is obtained through the use of this technique over the traditional LMF algorithm.

An adaptive learning rate for RBFNN using time-domain feedback analysis.

A major limiting factor in the performance of MIMO-CDMA systems is multiple access interference (MAI) which can reduce the system capacity as well as bit error rate (BER). Thus, statistical characterization of MAI is vital in analyzing the performance of such systems. Since, the statistical analysis of MAI in MIMO-CDMA is quite involved especially in the presence of fading channels, existing works in literature employ suboptimal approaches to detect the subscriber without involving the need for MAI statistics such as successive interference cancellation (SIC) and parallel interference cancellation (PIC). To date, the exact characterization of multiple access interference in MIMO-CDMA is an unsolved problem. In this paper, we derive the expressions for the probability density function of MAI and MAI plus noise in MIMO-CDMA systems in the presence of both Rayleigh fading channels and additive Gaussian noise. Simulation results show that the theoretical predictions are very well substantiated.

A noise constrained least mean fourth (NCLMF) adaptive algorithm

The least-mean fourth (LMF) algorithm is best known for its fast convergence and low steady-state error especially in non-Gaussian noise environments. Recently, there has been a surge of interest in the LMF algorithm with different variants being proposed. The fact that different variable step-size least-mean square algorithms have shown to outperform its fixed step-size counterpart, a variable step-size least-mean fourth algorithm of the quotient form (VSSLMFQ) is proposed here. Therefore in this work, the proposed algorithm is analysed for its performance in the steady-state and it is shown to achieve a lower steady-state error then the traditional LMF algorithm. Finally, a number of computer simulations are carried out to substantiate the theoretical findings.