The classification of incomplete patterns is a very challenging task because the object (incomplete pattern) with different possible estimations of missing data may yield distinct classification results. The uncertainty (ambiguity) of classification is mainly caused by the lack of information included in the missing data. Because of this, the object becomes really difficult to classify in such case. A new prototype-based credal classification (PCC) method is proposed in this work to deal with incomplete patterns thanks to the belief functions framework used classically in the evidential reasoning approach. The class prototypes obtained by the training data are respectively used to estimate the missing data. Typically, in a c-class problem, one has to deal with c prototypes which yields c estimations of the missing data. The different estimated (recovered) objects based on each possible estimation of the missing data are then classified by a standard classifier and we can get at most c distinct classification results for an incomplete pattern. Because all these distinct classification results are potentially admissible, we propose to combine them altogether to obtain the final classification of the incomplete pattern. In our approach, we propose a new credal combination method for solving the classification problem which is able to characterize the inherent uncertainty due to the possible conflicting results delivered by the different estimations of missing data attributes. The incomplete patterns that are very difficult to classify in a specific class will be reasonably and automatically committed to some proper meta-classes by this new PCC method in order to reduce the misclassification rate. The effectiveness of this new PCC method is tested through three experiments with artificial and real data sets.