Abstract: A complex nonlinear exponential autoregressive (CNEAR)modeling approach for the closed shape boundary is suggested in this paper. A neural network structure is constructed to estimate the parameters of the CNEAR model. The nonlinear coefficient is easy to be determined by the network. The CNEAR parameters are invariant to translation, rotation around the origin, scale and choice of the starting point in tracing a boundary. The feature vector can be constituted by the CNEAR parameters. A complex neural network classifier is designed for learning and classification and its learning algorithm is also given. Experimental results show that high recognition rates can be achieved even for low-order models. It is also seen that the CNEAR model performs better than the complex autoregressive (CAR) model when shapes have random noise on the boundaries or have differentiating features at detailed levels.