Accurately selecting neutron signals and discriminating gamma signals from a mixed radiation field is a key research issue in neutron detection. This paper proposes a fractal spectrum discrimination approach by means of different spectral characteristics of neutrons and gamma rays. Figure of merit and average discriminant error ratio are used together to evaluate the discrimination effects. Different neutron and gamma signals with various noise and pulse pile-up are simulated according to real data in the literature. The proposed approach is compared with the digital charge integration and pulse gradient methods. It is found that the fractal approach exhibits the best discrimination performance, followed by the digital charge integration method and the pulse gradient method, respectively. The fractal spectrum approach is not sensitive to high frequency noise and pulse pile-up. This means that the proposed approach has superior performance for effective and efficient anti-noise and high discrimination in neutron detection.