Abstract: Conventional SIB(Square Integral Bispectra,SIB) methods for feature extraction have several shortcomings: Firstly, Previous studies have not discussed the influence about integral path number on recognition rate. Secondly,there are some negative-effect integral paths.To overcoming these disadvantages,we proposed a new algorithm based on Improved Bispectra and Time-domain Analysis.First of all, The performance curve of integral path and recognition is obtained by exoeriments.Then using the largest proportion of energy inteval algorithm to remove low-contribution and negative-effect bispectrum values. Finally, the improved SIB and parameters significant for classification of the received signal formed the identification feature vector, and SVM(Support Vector Machine,SVM) was used to realize the individual identification. Experiment results show that the method is able to classify the same model transmitter with an accuracy rate of no less than 95% under the environment of lower SNR.