We present a framework for estimating pedestrian demand within a train station. It takes into account ridership data, and various direct and indirect indicators of demand. Such indicators may include link flow counts, density measurements, survey data, historical, or other information. The problem is considered in discrete time and at the aggregate level, i.e., for groups of pedestrians associated with the same origin-destination pair and departure time interval. The formulation is probabilistic, allowing to consider the stochasticity of demand. A key element is the use of the train timetable, and in particular of train arrival times, to capture demand peaks. A case study analysis of a Swiss train station underlines the practical applicability of the proposed framework. Compared to a classical estimator that ignores the notion of a train timetable, the gain in accuracy in terms of root-mean-square error is between 20% and 50%. More importantly, the incorporation of the train schedule allows for prediction when little or no data besides the timetable and ridership information is available.