This paper considers a fuzzy perceptron that has the same topological structure as the conventional linear perceptron. A learning algorithm based on a fuzzy δ rule is proposed for this fuzzy perceptron. The inner operations involved in the working process of this fuzzy perceptron are based on the max-min logical operations rather than conventional multiplication and summation, etc. The initial values of the network weights are fixed as 1. It is shown that each network weight is non-increasing in the training process and remains unchanged once it is less than 0.5. The learning algorithm has an advantage, as proved in this paper, that it converges in a finite number of steps if the training patterns are fuzzily separable. This result generalizes a corresponding classical result for conventional linear perceptrons. Some numerical experiments for the learning algorithm are provided to support our theoretical findings.