Variable leaky steepest descent algorithm for autoregressive fading estimation in OFDM systems
Introduction
Channel estimation is a crucial task in signal processing and in the new generation of wireless communications (see [1], [2], [3], [4], [5]). Inter-symbol interference (ISI) is avoided by using orthogonal frequency division multiplexing (OFDM). The frequency selective multipath channel can be converted to a set of parallel flat fading channels by using OFDM systems [5]. In orthogonal frequency division multiplexing (OFDM) systems carrier frequency offset (CFO) is a major problem in achieving orthogonality between subcarriers and introduces inter-carrier interference (ICI) [6], [7], [8] between subcarriers. CFO estimation scheme targeted for OFDM systems, using the subspace method, has been proposed in [9]. Many papers considered channel estimation in the presence of the CFO, that's why in this paper, we focused on autoregressive channel parameters estimation by an adaptive algorithm. For m-th subcarrier in OFDM system, the received signal, using a training sequence, can be expressed as where while M is total number of subcarriers, is an additive zero-mean white noise process, and is a Rayleigh flat fading which can be modeled as an autoregressive (AR) process [10], [11] given by
The driving noise, , is a zero-mean white noise process, and is the vector of AR parameters. The order of AR model, p, is assumed to be known. The noise processes and are assumed to be orthogonal, i.e. (), where E is the expectation operator and denotes conjugate transpose (). The variances of driving noise and observation noise are , and respectively. The autocorrelation function (ACF) for the fading process is given by [12] where while Q is number of autocorrelation lags, is the zero order Bessel function of the first kind, is the normalized Doppler spread, and is the variance of fading process. By taking the Fourier transform of , the power spectrum can be obtained
The AR modeling for the fading process can be accomplished by using the well-known Yule-Walker equations. We can select a proper AR order for fading channel by using AR order selection methods [13]. Then the AR model coefficients can be estimated by Yule-Walker equations.
In this paper, it is aimed to estimate the channel parameters , , and from the received signal samples i.e. () where N is the number of data samples.
From the above discussions, we see that the fading estimation in OFDM systems can be considered as a problem of AR estimation in noise. Many researchers have dealt with this problem (see [14], [15], [16], [17], [18], [19], [20], [21]). In the improved least squares (ILS)-based methods presented in [14], [15], [16], the inverse of the covariance matrix of the observations must be computed, which is a critical drawback for implementing these algorithms in terms of computational complexity and numerical instability. In [17] a computationally simple least mean squares (LMS) algorithm is devised for unbiased AR modeling in the presence of white noise, assuming that the observation noise is known; however, in fading estimation, the observation noise variance is unknown and we must estimate the observation noise variance. The method presented in [18] is based on the eigenanalysis of the covariance matrix of observations, which is difficult for online implementation. In [20], an errors-in-variables (EIV)-based method is presented. The method includes searching the noise variances whose noise compensated covariance matrix of observations is positive semi-definite. In spite of a good accuracy, the method has a high computational complexity. In [21], authors proposed vector process disturbed by an additive white noise. When this process, which can be corresponding to a mobile fading channel, is modeled by a multivariate autoregressive (M-AR) process, optimal filters such as Kalman filter can be used for prediction or estimation from noisy observations. Investigating the relevance of a structure based on two cross-coupled H∞ filters for the joint estimation of time-varying frequency-flat Rayleigh fading channel and its AR parameters proposed in [22]. In [23] authors addressed the problem of estimating a channel supposed to follow a Rayleigh model with Jakes' Doppler spectrum using a second order autoregressive model with Kalman filter and using a Minimization of Asymptotic Variance (MAV) criterion to improve the estimation performance.
In this paper, we present an autoregressive channel estimation scheme based on a variable leaky steepest descent (VLSD) algorithm for Rayleigh flat fading in OFDM systems. We show that the leak parameter in the VLSD should be considered equal to the observation noise variance. A computationally simple method for estimating the observation noise variance or the leak parameter is presented. The convergence analysis of the proposed adaptive algorithm is also performed and a stability criterion is derived.
Finally, given the AR parameters estimates, the fading process is estimated using Kalman filter and evaluated the Bit Error Rate (BER) performance in AWGN, Rayleigh fading channel before and after elimination of fading effect. The results are derived for BPSK modulation. In the simulation study, the performance of the proposed method is compared with the other existing methods.
This paper is organized as follows: Section 2 introduces the OFDM system and the AR modeling for fading processes. In Section 3, we derive the proposed VLSD algorithm for AR estimation in white noise. Section 4 presents the simulations to evaluate and compare the performance of the proposed method with the other existing methods. Finally, some conclusions are given in Section 5.
Section snippets
System model
The transmitted signal at time n in OFDM systems is given by
The sequence is a set of training symbols modulated by OFDM subcarrier and N is the total number of subcarriers which is greater than Z, while Z is length of training symbols. We assume that represents the zero mean uncorrelated sequences with .
The received signal at time n can be expressed as where shows the frequency selective fading
Derivation of the proposed AR estimation in noise
In this section, two cases are considered. In case 1, we assume that the observation noise variance is known and we derive an update rule for estimating the AR parameters. Reversely, in case 2, we assume that the AR parameters are known and we derive an estimation method for the observation noise variance. In subsections 3.3 and 3.4, the proposed algorithm and its performance are discussed, respectively. Kalman filter for estimating the fading process will be discussed in subsection 3.5.
Simulation results
In this section, a computer simulation example has been carried out to evaluate and compare the performance of the proposed method with the steepest descent (SD) method [24], and an ILS-based method from [16] (called Newton algorithm in this paper, for brevity).
Simulation parameters in an OFDM system are set as . Signal to noise ratio is defined by . We tune to reach a desired SNR. Using the OFDM parameters, and setting the AR order to be
Conclusions
Using the OFDM system, the frequency selective multipath channel can change to some parallel flat fading channels. Each channel can be modeled by a Rayleigh Autoregressive (AR) process. To estimate the channel parameters, the problem of AR modeling in presence of noise is considered. A new adaptive algorithm based on variable leaky steepest descent is proposed to yield an unbiased estimation of AR parameters. Using the AR model estimates, the fading process is estimated by exploiting Kalman
CRediT authorship contribution statement
Mohammad Ghanavati Mohammadi: Investigation, Resources, Software, Writing – original draft. Alimorad Mahmoudi: Conceptualization, Validation, Writing – review & editing. Abdolnabi Kosarian: Writing – review & editing.
Declaration of Competing Interest
The authors declare that they have no known competing financial interest or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgements
This study was financially supported by the Research Council of Shahid Chamran University of Ahvaz (Grant Number: SCU. EE98. 225).
Mohammad Ghanavati Mohammadi was born in Ahvaz, Iran, in 1986. He received the BSc degree in electrical Engineering in 2010. He obtained MSc degree in Communications Engineering from Shahid Chamran University of Ahvaz, Iran, in 2019. His research interests include signal processing, channel estimation and adaptive filters.
References (29)
- et al.
On the use of first-order autoregressive modeling for Rayleigh flat fading channel estimation with Kalman filter
Signal Process.
(2012) - et al.
Parameter estimation of autoregressive signals in presence of colored ar (1) noise as a quadratic eigenvalue problem
Signal Process.
(2012) - et al.
Inverse filtering based method for estimation of noisy autoregressive signals
Signal Process.
(2011) - et al.
Novel parameter estimation of autoregressive signals in the presence of noise
Automatica
(2015) - et al.
Second-order autoregressive model-based Kalman filter for the estimation of a slow fading channel described by the Clarke model: optimal tuning and interpretation
Digit. Signal Process.
(2019) - et al.
Adaptive signal processing for adaptive control
IFAC Proc. Vol.
(1983) - et al.
Channel estimation for 3d mimo system based on los/nlos identification
IET Commun.
(2019) - et al.
Robust multipath channel estimation in the presence of impulsive noise
IET Commun.
(2017) - et al.
Channel estimation for orthogonal frequency division multiplexing uplinks in time-varying channels
IET Commun.
(2014) - et al.
Estimation of carrier and channel parameters in time-selective fading channels
IET Commun.
(2015)
Joint data qr-detection and Kalman estimation for ofdm time-varying Rayleigh channel complex gains
IEEE Trans. Commun.
Analysis and mitigation of ici due to gain adjustment in ofdm systems
IEEE Access
Ici self-cancellation with cosine windowing in ofdm transmitters over fast time-varying channels
IEEE Trans. Wirel. Commun.
Blind channel shortening for asynchronous sc-ifdma systems with cfos
IEEE Trans. Wirel. Commun.
Cited by (2)
Combined Diffusion Adaptation on Adaptive Leaky Criterion and Orthogonal Gradient Algorithm
2023, 2023 20th International Conference on Electrical Engineering/Electronics, Computer, Telecommunications and Information Technology, ECTI-CON 2023Early Stopping Criterion for Recursive Least Squares Training of Behavioural Models
2022, Wireless Personal Communications
Mohammad Ghanavati Mohammadi was born in Ahvaz, Iran, in 1986. He received the BSc degree in electrical Engineering in 2010. He obtained MSc degree in Communications Engineering from Shahid Chamran University of Ahvaz, Iran, in 2019. His research interests include signal processing, channel estimation and adaptive filters.
Alimorad Mahmoudi received the B.S., M.S., and Ph.D. degrees in Electrical Engineering from Shiraz University, Shiraz, Iran, in 2003, 2005, and 2010, respectively. Since 2010, he has been an Assistant Professor at the Department of Electrical Engineering at Shahid Chamran University of Ahvaz, Ahvaz, Iran. His research activities are focused on signal processing for communications systems.
Abdolnabi Kosarian was born in Behbahan, Khouzestan, Iran. He received his Ph.D. degree in Electronic Engineering from University of Surrey, Surrey, UK, in 1998. He is currently a professor of electronic engineering at Shahid Chamran University of Ahvaz, Ahvaz, Iran. His research interests include solid state electronics, semiconductor devices and solar cell technology.