A new neural network architecture is introduced for incremental learning fuzzy measure for Choquet integrals based fusion system. The proposed approach differs from other determining fuzzy measure methods by its suitability to incremental learning fuzzy measures. The fuzzy measure is treated as the weights of the neural networks. An additional neural network is introduced to accommodate new data when new samples arrive, including examples that correspond to previously unseen classes. Furthermore, the algorithm does not require access to previously used data during subsequent incremental learning sessions, but preserve the center of a batch of samples. At the same time, it does not forget previously acquired knowledge. The outputs of the resulting classifier fusion systems are combined using a weighted majority procedure where the weights are dependent on the performance of the fusion system and the distance of testing samples and the centers of each batch. We present simulation results on several benchmark datasets as well as a real-world classification task. Initial results indicate that the proposed algorithm works rather well in practice.