Similarity Based on Power Spectrum Analysis ?

The predecessors calculated the text similarity established on quantitative calculation. The paper propose a new method to distinguish text similarity on the basis of power spectrum, mainly expressed the text using the form of spectrum diagram. It is not difficult to imagine that the signals of brain are closely linked with writing process. The method combined with this idea to build text modeling and set pulsing signal function which is converted to the frequency domain achieving the power spectrum by MATLAB. The specific detail is get power spectrum from economic field to build spectral library using the method of power spectrum matching algorithm to judge whether the test text belong to the economic field or not. Related experimental results show that this method is easy to implement and the precision is improved greatly.