The invention provides a symmetric binary tree decomposition method for optimizing an FFT (fast fourier transform) mixed base algorithm. The symmetric binary tree decomposition includes the steps: constantly decomposing given-point FFT into a combination of a plurality of small-point FFT according to the symmetric method for carrying out the iterative cycle decomposition process of the given-pointFFT; for the base cycle replacement process, performing cyclic iteration switch on bases corresponding to two neighboring sublevels in a base sequence generated from first decomposition; completing optimization of the FFT mixed base algorithm in combining with results of multiple iterative symmetric decomposition and base cycle replacement. With the method, the mixed base algorithm for any given2 integer power points FFT can be optimized rapidly and efficiently, multiplication complexity related to rotation factors can b reduced effectively, and FFT overall operational efficiency is improved.