A neural network approach to the partial shape classification problem

A neural network approach to the partial shape classification problem is derived. Although neural networks are robust static pattern classifiers, they are generally not effective in classifying patterns with inherent time variations. In order to compensate for time variations resulting from random partial occlusion, a nonlinear alignment stage is introduced at the neural net output. In formulating the nonlinear alignment stage, a similarity measure between an input and the neural net outputs is defined. The resulting classifier is capable of tolerating high degrees of random noise and random occlusion in shapes