For recognizing the signals of pulsars and improving the effect of recognition,non-linear and non-gaussian characteristics of the average pulse profiles of pulsar signals were analyzed by using bispectra technology.Then one-dimension selected line spectra algorithm for extracting pulsar signals' characteristics was proposed.The algorithm makes use of feature core and domain to represent the generality of bispectrum and its distribution domain.And the variance of signal feature vector and the distance between two kinds of signal feature vectors were defined,based on which one-dimension line spectral vector with best interclass separability was redefined.The results of recognition experiments show that the algorithm of one-dimension selected line spectra is suitable for extracting pulsar signals' characteristics.Compared with the selected bispectra,the selected line spectrum has the maximum interclass separability.