Abstract: Mining splice features is the foundation of the algorithm of splice sites identification. It is important to extract useful features in frequency domain. Extract splice site features based on fast Fourier transform. It can transform the information from spatial domain to frequency domain and use the information in frequency domain as the new splice feature. In order to make a comparison, position features and statistic features were constructed. The experiments' results show that the splice feature extracted by our way has better effect and it also indicates that signal processing methods can be used in bioinformatics efficiently.