Study On Blind Per Tone Equalization Algorithm Based On Twin Support Vector Regressor For OFDM Systems

Orthogonal frequency division multiplexing(OFDM) technology can effectively combat frequency selective fading and improve the frequency utilization.For OFDM systems,the insertion of cyclic prefix can mitigate inter-symbol interference(ISI) and inter-carrier interference(ICI) to some extent. If the biggest channel impulse respond is longer than the CP, the performance of system will degrade seriously. In this case, channel equalization technology is needed to compensate the influence of channel and eliminate ISI and ICI. Traditional time-domain equalization(TEQ) method equalizes all tones in a combined fashion. For a certain tone, this equilibrium structure is not the best because different tones undergo different fading. Per tone equalization(PTEQ) method insert a frequency-domain equalizer(FEQ) in each tone. This method equalizes every tone independently so that the system can get a better performance.Blind equalization as a kind of adaptive equalization technology that needn’t the training sequence and just use the statistical property of original sequence to restore the transmit signals. It can effectively improve the frequency utilization. In this article, we mainly study the blind per tone equalization of OFDM system.This article introduced the basic principle of OFDM technology and blind equalization firstly,analyzed the structure and principle of per tone equalizer, and outlined two traditional blind per tone equalization algorithms of OFDM system: blind per tone constant modulus equalization algorithm(PT-CMA) and blind per tone multimodulus equalization algorithm(PT-MMA) of OFDM system. Then the article detailed the blind per tone equalization algorithm based on support vector regressor(PT-SVR) of OFDM system. Based on above discussion, we first proposed the blind per tone equalization algorithm based on twin support vector regressor(PT-TSVR) of OFDM systems. The algorithm combines higher-order statistics(HOS)method applied to blind equalization with the conventional cost function of the TSVR: Satos’ s and Godard’s error functions contained in the penalty term of the TSVR, and the equalizer coefficientsare obtained quickly by means of the iterative re-weighted least square(IRWLS) algorithm.Moreover, dual mode equalization scheme is used to against the problem that HOS blind error functions would present a high residual error for multilevel signals. Once the eye is opened by the PT-TSVR blind equalization algorithm, system switches to decision directed(DD) algorithm.Simulation experiments show that the bit error rate(BER) of PT-TSVR blind equalization algorithm is lower than traditional PT-CMA and PT-MMA blind algorithms obviously. Compared with PT-SVR blind equalization algorithm,the BER of PT-TSVR also improves. PT-TSVR+DD dual mode equalization scheme can reduce the BER further.