In this work the recently developed fuzzy logic-based adaptive Kalman filter (FL-AKF) is integrated into a neurofuzzy network structure to perform system identification and state estimation of unknown nonlinear systems. This approach, referred to as neurofuzzy adaptive Kalman filter, uses the error signal in the identification process as the measurement noise signal for the FL-AKF in order to estimate the modelling error at the same time in which system identification is performed by the neurofuzzy network. This has a stabilisation effect during the training process when noise is present in the training data. A simulated example is presented to validate the effectiveness of the proposed approach