关键词: 峭度, 对比函数, 参考独立成分分析, 信噪比

Abstract: One-unit ICA-R is an efficient method utilizing prior information to extract an expected source signal.Kurtosis is a classical measure of non-Gaussianity of random variable.Based on constrained independent component analysis，a fast algorithm for one-unit ICA-R is proposed when absolute value of kurtosis is considered as contrast function in this paper.A better initial value is provided to accelerate convergence based on convergent characteristic of the algorithm.At last，computer simulations verify the validity of the proposed algorithm and indicate that the better initial value can improve convergence indeed.

Key words: kurtosis, contrast function, Independent Component Analysis with Reference（ICA-R）, Signal to Noise Ratio（SNR）