一種峭度依賴的參數(shù)自適應(yīng)盲分離算法
A Parameter Kurtosis-Dependent Flexible BSS Algorithm
-
摘要: 針對超高斯與亞高斯混合信源分離算法上存在的不足,該文提出一種峭度依賴的參數(shù)自適應(yīng)盲分離算法。該算法用加權(quán)雙高斯模型估計(jì)超高斯與亞高斯信源分布,在自然梯度框架下,依據(jù)峭度實(shí)現(xiàn)模型參數(shù)自適應(yīng)。通過使用混合圖像對其進(jìn)行驗(yàn)證,實(shí)驗(yàn)表明該算法不僅可以有效實(shí)現(xiàn)超高斯與亞高斯混合信源的分離,而且比已有算法具有更好的分離和收斂性能。
-
關(guān)鍵詞:
- 信號處理; 盲分離; 加權(quán); 雙高斯; 超高斯; 亞高斯
Abstract: To overcome some shortcomings of existing algorithms which separate the mixture of super- and sub-gaussian sources, a parameter kurtosis-dependent flexible Blind Source Separation (BBS) algorithm is proposed. A weighed double Gaussian model is proposed to estimate super-Gaussian and sub-Gaussian probability density. In the framework of natural gradient, model parameter is calculated online by kurtosis. Applied to images mixing, experiment shows the proposed algorithm can successfully separate the mixture of super- and sub-gaussian images. Meanwhile experiment also shows that the proposed algorithm has better performance and convergence than existing algorithms. -
Amari S I. Natural gradient works efficiently in Learning[J].Neural Computation.1998, 10(2):251-276[2]Cardoso J F. Blind signal separation: Statistical principles[J].Proc. IEEE.1998, 86(10):2009-2025[3]Boscolo R, Vwani H P. Independent component analysis based on nonparametric density estimation[J].IEEE Trans. on Neural Networks.2004, 15(1):55-65[4]Vlassis N, Motomura Y. Efficient source adaptivity in independent component analysis[J].IEEE Trans. on Neural Networks.2001, 12(3):559-565[5]Lee T W, Girolami M, Sejnowski T J. Independent component analysis using an extended informax algorithm for mixed sub-gaussian and super-gaussian sources. Neural Computation, 1999, 11(2): 409-433.[6]Choi S, Cichocki A, Amari S. Flexible independent component analysis. IEEE Workshop on Neural Networks for Signal Processing, Cambridge, UK, 1998: 83-92.[7]Hyvarinen A, Karhunen J, Oja E. Independent Component Analysis. New York: John Wiley, 2001: 203-208. -
計(jì)量
- 文章訪問數(shù): 2130
- HTML全文瀏覽量: 94
- PDF下載量: 655
- 被引次數(shù): 0