Blind Linear Interference Suppression Based on Reduced-Rank Least-Squares Constrained Constant Modulus Design for DS-CDMA Systems

In this paper we present a blind interference suppression technique for DS-CDMA systems based on a reduced-rank decomposition of a code-constrained constant modulus design criterion. We describe a least-squares (LS) type design criterion for blind linear detectors using the constrained optimization of the constant modulus cost function subject to code constraints. Based on this design approach and using the Lanczos algorithm, a multistage decomposition in the Krylov subspace is devised for blind reduced-rank parameter estimation. A computationally efficient blind reduced-rank LS type algorithm is also developed and compared with existing methods. Numerical results show that the proposed full-rank and reduced-rank techniques outperform existing methods for the linear suppression of multi-access and intersymbol interference in DS-CDMA systems.