Computationally efficient Toeplitz-constrained blind equalization based on independence

Abstract

Bussgang algorithms for blind equalization (BE) are well-known techniques that use known constellation properties of the source symbols. BE has also been based on independent component analysis (ICA) methods, where independence of the symbols is exploited. When channel outputs are processed in blocks, as in the ICA algorithms, extra computational cost is incurred for matrix operations. In this paper we propose a computationally efficient Toeplitz-constrained BE algorithm based on independence. We show how the FFT can be applied for computation efficiency, and also how blockwise sample cross-correlation computations can be efficiently approximated. For sources with independent I/Q components this constraint can be imposed for phase recovery. Our proposed algorithms yield relatively fast convergence without high computational cost.