Mohammad Shukri Ahmad

Recursive inverse adaptive filtering algorithm

In this paper, a new FIR adaptive filtering algorithm is introduced. This algorithm is based on the Quasi-Newton (QN) optimization algorithm. The approach uses a variable step-size in the coefficient update equation that leads to an improved performance. The simulation results show that the algorithm has very similar performance to the Robust Recursive Least Squares Algorithm (RRLS) while performing better than the Transform Domain LMS with Variable Step-Size (TDVSS) in stationary environments. The algorithm is tested in Additive White Gaussian Noise (AWGN) and Correlated Noise environments.

An efficient Recursive Inverse adaptive filtering algorithm for channel equalization

Recursive Inverse (RI) adaptive filtering algorithm which uses a variable step-size and the instantaneous value of the autocorrelation matrix in the coefficient update equation was proposed in [1]. The algorithm was shown to have a higher performance compared with the RLS and RRLS algorithms. In this paper, a more efficient version with lower computational complexity is presented. The performance of the algorithm has been tested in a channel equalization setting and compared with those of the Recursive Least Squares (RLS) and Stabilized Fast Transversal Recursive Least Squares (SFTRLS) algorithms in Additive White Gaussian Noise (AWGN), Additive Correlated Gaussian Noise (ACGN), Additive White Impulsive Noise (AWIN) and Additive Correlated Impulsive Noise (ACIN) environments. Simulation results show that the Fast RI algorithm performs better than RLS and requires less computations. Additionally, the performance of the Fast RI algorithm is shown to be superior to that of the SFTRLS algorithm under the same conditions.

A 2-D recursive inverse adaptive algorithm

In this paper, a 2-D form of the recently proposed recursive inverse (RI) adaptive algorithm is introduced. The filter coefficients can be updated along both the horizontal and vertical directions on a 2-D plane. The proposed approach uses a variable step size and avoids the use of the inverse autocorrelation matrix in the coefficient update equation, which leads to an improved and more stable performance. Performance of the 2-D RI algorithm is compared to that of the 2-D RLS algorithm in an image deconvolution and an adaptive line enhancer problem settings. The simulation results show that the proposed 2-D RI algorithm leads to an improved performance compared to that of the 2-D RLS algorithm.

Robust Recursive Inverse Adaptive Algorithm in Impulsive Noise

The recently proposed Recursive Inverse (RI) algorithm has shown a significant performance improvement compared to that of the Recursive Least Squares (RLS) algorithm, in various noise environments. However, both algorithms fail to converge in certain impulsive noise environments, especially if the Signal-to-Noise Ratio (SNR) is low. In this paper, a Robust RI algorithm is proposed. Analytical results show that robustness against impulsive noise is achieved by choosing the weights on the basis of the L1 norms of the autocorrelation matrix and the cross-correlation vector. Simulation results confirm that the proposed algorithm provides an improved performance, with a reduction in computational complexity, compared to those of the RLS and the Robust RLS in white and correlated impulsive noise.

The effect of the forgetting factor on the RI adaptive algorithm in system identification

The recently proposed Recursive Inverse (RI) algorithm was shown to have a similar mean-square-error (mse) performance as the Recursive-Least-Squares (RLS) algorithm with reduced complexity. The selection of the forgetting factor has a significant influence on the performance of the RLS algorithm. The value of the forgetting factor leads to a tradeoff between the stability and the tracking ability. In a system identification setting, both the filter length and a leakage phenomenon affect the selection of the forgetting factor. In this paper, we first analytically show that this leakage phenomenon and the filter length have much less influence on the performance of the RI algorithm. Simulation results, in a system identification setting, validate the theoretical results.

Recursive inverse adaptive filter with second order estimation of autocorrelation matrix

The recently proposed Recursive Inverse (RI) Adaptive Filtering algorithm uses a variable step-size and the first order recursive estimation of the correlation matrices in the coefficient update equation which lead to an improved performance. In this paper, a new FIR adaptive filtering algorithm is introduced. This algorithm uses the second order recursive estimation of the correlation matrices in the coefficient update equation which leads to an improved performance over the RI algorithm. The simulation results show that the algorithm outperforms the Transform Domain LMS with Variable Step-Size (TDVSS), the RI and the RLS algorithms in stationary environments. The performance of the algorithms is tested in Additive White Gaussian Noise (AWGN) and Correlated Noise environments.

A two dimensional recursive inverse adaptive filtering algorithm

In this paper, a 2-D form of the recently proposed Recursive Inverse (RI) algorithm is introduced. The filter coefficients can be updated both along the horizontal and vertical directions on a 2-D plane. The proposed approach uses a variable step-size and avoids the use of the inverse autocorrelation matrix in the coefficient update equation, which leads to an improved and more stable performance. In order to compare with the well-known 2-D RLS adaptive filtering algorithm, it is applied to an image deconvolution problem. Simulation results show that the proposed 2-D algorithm leads to improved performance compared to that of the 2-D RLS algorithm.

A fast-implemented recursive inverse adaptive filtering algorithm

The recently proposed Recursive Inverse (RI) adaptive algorithm [1] has shown improved performance in channel equalization and system identification settings. Although its computational complexity is lower than those of the RLS and Robust RLS algorithms, its computational complexity can be reduced further. A fast implementation method is applied in this paper to decrease its computational complexity. The performance of the fast implemented RI algorithm is compared to those of the Variable Step-Size LMS (VSSLMS), Discrete Cosine Transform LMS (DCTLMS) and Recursive-Least-Squares (RLS) algorithms in Additive White Gaussian Noise (AWGN), Additive Correlated Gaussian Noise (ACGN), Additive White Impulsive Noise (AWIN) and Additive Correlated Impulsive Noise (ACIN) environments in a noise cancellation setting. Simulation results show that the Fast RI algorithm performs better than the VSSLMS and DCTLMS algorithms. Its performance is the same as in the RLS algorithm with a considerable reduction in complexity.

Detection in correlated impulsive noise channels using Frequency-Response-Shaped adaptive filtering

This paper investigates the performance of an adaptive filter, (Frequency-Response-Shaped Least Mean Square (FRS-LMS) algorithm) for canceling impulsive components when the nominal process (or background noise) is a correlated, possibly nonstationary, Gaussian process. The performance of the algorithm in estimating a BPSK signal corrupted by a white and correlated impulsive noise is investigated. The algorithm does not require a priori knowledge about the noise parameters, but requires knowledge of the signal frequency which can easily be estimated from its periodogram. The performance of the FRS-LMS is compared to that of the conventional LMS, the Leaky-LMS (L-LMS), and the Modified Leaky LMS (ML-LMS) algorithms in terms of Mean Square Error (MSE), convergence speed and Bit-Error-Rate (BER). The results indicate that the FRS-LMS algorithm performs approximately twice as better than the LMS and L-LMS algorithms in white impulsive noise environments, while the ML-LMS algorithm fails to converge. Also, it provides superior MSE and BER performance in correlated impulsive noise environments, while the other algorithms fail to converge. The performance gain is due to the frequency shaping and the outlier reduction properties of the algorithm.

Adaptive channel estimation using frequency-response-shaped LMS algorithm in DS-CDMA systems

An adaptive least mean square (LMS) type channel estimation algorithm is proposed. The proposed algorithm employs the frequency-response-shaped least mean square (FRS-LMS) algorithm which outperforms the LMS and the modified leaky LMS algorithms in a fading channel with correlated Gaussian noise. The algorithm is based on shaping the frequency response of the transversal filter by inclusion of a leakage factor in matrix form. The FRS-LMS algorithm converges to a lower MSE compared with the modified leaky LMS in correlated Gaussian noise leading to a substantial decrease in bit error rate for direct sequence code division multiple access (DS-CDMA) systems. In a fading channel with AWGN, the proposed algorithm has a lower performance gain relative to the modified leaky LMS algorithm but provides a significant performance advantage over the standard LMS algorithm of the system.