The invention discloses a multi-antenna OFDMA signal decoding method based on rapid and independent component analysis. The method comprises the following steps: step 1, establishing a signal model for meeting the requirement of a rapid and independent component analysis algorithm on an input signal; step 2, conducting PCA whitening on an observation vector under Gaussian noise; step 3, separating a plurality of user information modulated on OFDMA subcarriers without knowing any information about the subcarriers; and step 4, performing subcarrier sequence estimation and OFDMA user identification. According to the method, mathematical statistical modeling is conducted on multi-antenna OFDMA signals based on Fast-ICA, and blind channel estimation and decoding are also performed, so that the signal energy is increased without removing a cyclic prefix, and the decoding performance is improved. Through adoption of the method, timing is not needed at a receiving end, blind separation and decoding are conducted on the OFDMA signals, and the decoding performance is good.