In this paper, we propose a new chirp ranging system using neural networks to improve the ranging accuracy. In the proposed system, neural networks are employed for two purposes; one is to filter out noises contained in the measured signals that include multipath signals with variable time delay and amplitude, and another is to exactly detect a peak point for calculating time-of-flight. We conducted simulations to examine the validity of the proposed system. It will be shown that the ranging precision can be significantly improved by the proposed chirp ranging system.