To discriminate orthogonal frequency division multiplexing(OFDM) from single carrier signals, this paper proposes three new feature parameters based on spectral analysis. Since the edge of power spectrum of the OFDM is sharp, by statistical analysis and curve fitting the normalized power spectral, three new feature parameters are extracted, and the recognition simulation of OFDM and 9 kinds of single carrier signals under the additive white Gaussian noise (AWGN) is carried out. The method has lots of advantages, such as the feature parameters could be extracted easily and recognition accuracy rates are also very high.