We present an approach allowing to automatically ex- tract a suitable set of soft output classifiers and to aggre- gate them to provide a global decision using the Choquet integral. This approach relies on two key points. A learning algorithm based on a 2-class model is performed to define a new set of decisions rules assuming to be experts dedicated to recognize one class from another one. All the associated capacities are aggregated again at a high level to recog- nize symbols. The second is a selection scheme that dis- cards weak or redundant decision rules, keeping only the most relevant subset. An experimental study, based on real world data, is then described. It analyzes the improvements achieve by these points first when used independently, then when combined together.