Abstract: The main types of blind signal processing algorithm are batch algorithm and adaptive agorithm. Combined with batch algorithm and adaptive agorithm, the fast independent component analysis algorithm for speech signal blind separation processing is presented in this paper. Through the comprehensive experiments, the results show that Fast ICA algorithm has good signal separation efficiency from the signal waveforms and spectrums before and after separation and the main evaluation parameters. Fast ICA algorithm has better separation efficiency than the joint approximative diagonalization of eigenmatrix algorithm and natural gradient algorithm.