A Study Of Blind Equalization Algorithm Based On Second-order Statistics

Equalization technology is an effective method which can remove inter-symbol interference (ISI) in digital communication. Blind equalization technology is a very popular research topic because it can adaptively equalize without training sequence. Blind equalization algorithm based on cyclostationarity was presented and developed in the middle of 90's. Among them, blind equalization based on SOS (second order statistics) is most popular. SOS convey not only amplitude but also complete phase information, so system can be equalized only by using output signal.(1) This paper researches blind equalization rulers, analyzes the knowledge of SOS and summarizes two kinds of algorithms based on SOS. The necessity and feasibility of applying this kind of algorithm are also illustrated.(2) This paper sets up Single-input Multiple-output (SIMO) model and describes the subspace method of SIMO systems. The process of signal transmission through the equalizer is simulated. This method can identify the transfer function of the system only using SOS by sampling multiple outputs or oversampling a single output. It lays an important foundation for blind equalization and blind identification.(3) ZF blind equalizer is called perfect equalizer when noise is ignored. But when taking noise into account the equalization properties would be badly influenced. In order to overcome these drawbacks a new direct blind MMSE equalization algorithm based on SOS is derived in this paper. According to MMSE criterion, the algorithm could use second-order cyclostationarity to equalize channel directly without identification.(4) The Bit Error Rate (BER) performance of the new algorithm is theoretically analyzed and simulated. The comparisons with blind equalizations based on subspace method and CMA are shown through computer simulations. Then it is found that the proposed algorithm has the advantages in low complexity, and will not drop into local optimization. Besides these, the algorithm turns out to be more robust against channel order mismatch for the direct method doesn't need identification. What's more, the algorithm attains an excellent BER performance.