Lower bounds on the information rate of intersymbol interference channels based on the Ungerboeck observation model

We consider the problem of lower bounding the information rate of intersymbol interference channels via Monte Carlo algorithms. We adopt a novel approach based on the Ungerboeck observation model, unlike the existing ones that are all based on the Forney model. The two approaches, yet equivalent in the case of full-complexity detection, lead to different results in the case of reduced-complexity detection, which is the only viable option when the channel memory is large. The proposed approach significantly outperforms the existing ones for all practical values of the signal-to-noise ratio.