On estimation-based iterative detection of one-dimensional and two-dimensional ISI channels

Detection algorithms working directly on discrete alphabet for one-dimensional long intersymbol interference (ISI) channels and two-dimensional ISI channels usually have exponentially high complexities. This paper proposes new estimation-based iterative detection algorithms working on continuous alphabets. These algorithms achieve near-optimal error performance with low complexity. They are based on maximization of the likelihood function in the continuous domain using the gradient descent (ascent) method with iterative application of constraints. The monotonicity is a desirable property for an iterative estimation algorithm in maximizing the likelihood function for designing new estimation-based detection algorithms.