Blind channel estimation using the second-order statistics: algorithms

Most second-order moment-based blind channel estimators belong to two categories: i) optimal correlation/spectral fitting techniques and ii) eigenstructure-based techniques. These two classes of algorithms have complementary advantages and disadvantages. In this paper, a new optimization criterion referred to as the joint optimization with subspace constraints (JOSC) is proposed to unify the two types of approaches. Based on this criterion, a new algorithm is developed to combine the strength of the two classes of blind channel estimators. Among a number of attractive features, the JOSC algorithm does not require the accurate detection of the channel order. When compared with existing eigenstructure-based techniques, the JOSC performs better, especially when the channel is close to being unidentifiable. When compared with correlation/spectral fitting schemes, the JOSC is less affected by the presence of local minima.