An Adaptive Channel Parameter Estimation Using QQ-plot Z. DJUROVIC * , I. KOVACEVIC ** , B. KOVACEVIC * * Faculty of Electrical Engineering, University of Belgrade, SERBIA ** High School of Electrical Engineering, SERBIA zdjurovic@etf.bg.ac.yu ; ivanak@vets.edu.yu; ,kovacevic_b@etf.bg.ac.yu Abstract. New algorithm for estimation of parameters of communication channel in the circumstances of existence of intensive impulse noise within measurement sequence is proposed in this paper. Proceeding from the theory of robust estimation, a simple, adaptive, practically applicable algorithm is derived that in the circumstances of contaminated normal distribution of measurement noise demonstrates high level of efficiency. QQ-plot technique is used as a framework for estimation of contaminated measurements distribution providing the algorithm adaptation. Application of proposed algorithm is broad, both in the field of wireless communications, equalization of transmitting channels, suppressing of noise and in modeling communication and control systems. Keywords: Channel identification, robustness, adaptation, QQ-plot. 1. Introduction In wireless communication systems, the transmitted signals usually experience fading which either attenuates the received power or causes dispersion. Therefore, it is necessary to obtain the fading channel information, i.e. the channel gains of different resolvable paths. However, the channel state estimation is unknown to both the transmitter and the receiver in many practical applications, thus necessitating channel estimation at the front end of the coherent receiver [1]. A variety of channel estimation algorithms have been developed, based primarily on one of two aspects of random variables: the distribution and the moments. When the distribution of the received signal conditioned on the channel state information is known, the maximum likelihood (ML) criterion can be applied to yield asymptotically optimal performance [2]. Such ML channel estimation algorithms are suitable for training-symbol based systems, in which a subset of the transmitted symbols is known to both the transmitter and receiver [3]. The main concern with such a ML channel estimation is the necessary amount of training data [4]. When only the information symbols are available, this is usually called blind channel estimation. In the recent past, a large number of blind channel estimation algorithms have been developed using moment estimation, particularly the second-order statistics or SOS [5]. Usually, based on SOS estimation, the subspace method and moment matching are applied. However, the subspace technique is suitable only for stationary channels, since it requires the signal subspace to be time-invariant. However, in many practical code division multiple-access (CDMA) systems long codes are employed, thus making the overall channel nonstationary. On the other hand, SOS-based estimation techniques are commonly recognized as the natural tools to be used in the presence of Gaussian noise. Research efforts on higher-order statistics (HOS) have led to the development of improved estimation algorithms for non-Gaussian environments. Important non- Gaussian impulsive processes are found in a variety of practical problems that include wireless communications and teletraffic. These processes can be efficiently modeled by heavy-tailed distributions with a huge variance, for which neither the classical SOS theory nor the theory of HOS are well-defined [6]. Additionally, in many applications one expects that channel estimation can be made adaptively so as to accommodate time-varying environments and system parameter variation. The recursive least- squares (RLS) algorithms may be one of the most interesting and powerful techniques that implement adaptive parameter estimation [7]. However, when an impulsive noise in the system output exists, the RLS algorithm usually fails to yield an unbiased estimate of the system parameters, thereby causing the performance of adaptive filtering to be significantly degenerated. The robust estimation theory represents a suitable tool to cope with an impulsive noise environment. In this paper the problem of robustified adaptive parameter channel estimation in the min-max robust estimation context is considered. WSEAS TRANSACTIONS on CIRCUITS AND SYSTEMS Z. Djurovic, I. Kovacevic, B. Kovacevic ISSN: 1109-2734 600 Issue 7, Volume 7, July 2008