Fuzzy Adaptive Blind Equalizer Using Extended Kalman Filter Based Adaptation Algorithm For Powerline Channel Wai Kit Wong and Heng Siong Lim Faculty of Engineering and Technology, Multimedia University, Jln Ayer Keroh Lama, 75450 Melaka, Malaysia. Tel:+60 06-2523258 Email:wkwong@mmu.edu.my;hslim@mmu.edu.my Abstract Fuzzy adaptive equalizer (FAE) is a knowledge based equalizer operating on linguistic variables. The advantages of using fuzzy logic adaptation scheme with respect to more traditional adaptation schemes in powerline communication system are the simplicity of the approach and the use of knowledge (fuzzy IF- THEN rules and input output pairs information) about the communication medium. This paper presents a new adaptive blind equalization method based on fuzzy logic for powerline channel. We introduce a new type of fuzzy adaptive blind equalizer (FABE) using extended Kalman filter (EKF) based adaptation algorithm for powerline channel equalization. The proposed blind equalizer for powerline channel has the following merits: It is new and simple in design, and it does not requires training sequence. In a changeable distorted powerline channel, data transmission is continuous and do not stop for training the equalizer. The performance of EKF-based FABE is compared with two other types of FABEs based on the recursive least squares (RLS) and the least mean squares (LMS) adaptation algorithm. The simulation results show that EKF-based FABE has faster convergent and lower steady state probability of error compared to the other two FABEs. The bit error rate (BER) of the EKF- based FABE is close to that of the optimal equalizer. I. INTRODUCTION The term equalizer or filter in powerline communication system is commonly referred to a device that is designed to perform signal processing operations by extracting information about a prescribed quantity of interest from noisy and intersymbol interfered data [1]. In conventional equalization process, initial acquisition of the equalizer’s parameters is usually accomplished using learning sequences transmitted periodically in time. However, for the sake of simplicity in system design and convenience in system implementation, it is sometimes desirable to let the receivers start up without the aid of the transmitter. For certain powerline communication system, it is impractical to utilize a training sequence of long duration and the system cost involved to train the equalizer with repeated transmission of a known sequence is typically high. Hence a brilliant way to overcome such problems is by applying blind equalization technique. Blind channel equalization, or more formally, unsupervised equalization, is a type of equalization technique that depends only on the received samples and assumptions on the input data to identify the channel [1]. The idea of blind equalization systems dates back to the pioneering works of Sato [2], [3] and Godard [4], [5]. The Sato algorithm for blind equalization was introduced originally to deal with one-dimensional multilevel pulse amplitude modulation (M-ary PAM) signals, which is more robust than a decision-directed algorithm [1]. Godard algorithm is considered to be the most successful among the Bussgang family of blind equalization algorithms, as demonstrated by the comparative studies done in [6] and [7]. The Godard algorithm is more robust than other Bussgang algorithm due to the fact that cost function used for its derivation is based solely on received signals’ amplitudes [8]. Under steady state conditions, blind equalizer using Godard algorithm attains a mean square error that is the lowest among all the Bussgang algorithm existed in the world nowadays [8]. In this paper, we modify on a class of Bussgang-type equalizers that employ the Godard algorithm. In particular, we have replaced the transversal filter in the conventional Bussgang- type equalizer with a fuzzy adaptive filter. The modified Bussgang type equalizers (i.e., fuzzy adaptive blind equalizer (FABE)) operate on the received signals sampled at the baud rate. Unlike most of the blind equalizers recently proposed, FABE is a knowledge based system operating on linguistic variables. Linguistic information (fuzzy IF-THEN rules) and numerical information (input-output pairs) can be combined into the blind equalizer. We further derive a FABE that uses an extended Kalman filter (EKF) adaptation algorithm [9] for equalization of powerline channel. Simulation results show that this type of FABE has faster convergent speed compared to both RLS-based FABE and LMS-based FABE. The paper is organized as follows: Section II briefly comments on powerline data transmission system using FABE. Section III describes the structure of the proposed FABE and the adaptation algorithm. Section IV contains some computer simulation results and Section V summarizes the work. II. POWERLINE DATA TRANSMISSION SYSTEM USING FABE The baseband powerline data transmission system using FABE is shown in Fig. 1. noise n(k) The channel includes the effects of the transmission filter, powerline communication (PLC) channel and reception filter. The + message signal s(k) CHANNEL Fuzzy Adaptive Filter g (f(x(k))) + Adaptation Algorithm f(x(k)) + - y(k) ê(k) x(k) equalizer output FABE, a modified Bussgang type equalizer Fig 1: Powerline data transmission with fuzzy adaptive blind equalization IV  449 142440469X/06/$20.00 ©2006 IEEE ICASSP 2006