臺灣博碩士論文加值系統

本篇論文針對迴旋式獨立來源混合的隱蔽式分離做深入的探討及研究。時間的迴旋結構根據假設是由多個混合矩陣所組成，其中每個矩陣對應到一段時間延遲；共同地轉換一部份連續的來源訊號，形成多個頻道的觀測值。當 τ=1 時，問題簡化成線性獨立成分分析。對任意的 τ>1，我們提出一個新的演算法，不但能估計不明的迴旋結構而且也能估計獨立來源。所提出的迴旋式獨立成分分析演算法是以最佳化核心估計，以及在平均場退火過程操作下的 leave-one-out approximation 為基礎。我們以人為的資料和兩隻麥克風所收錄的人聲和音樂來測試這個新的演算法。結果顯示對人為的資料而言，在給定及估計的迴旋結構之間的誤差明顯地減少；而且對兩隻麥克風所收錄的聲音，人聲和背景音樂也有不錯的分離成果。按照以往的經驗，所提出的新演算法對於迴旋式獨立來源混合的隱蔽式分離是有作用的。
關鍵字：迴旋式混合、獨立成分分析、隱蔽式分離、最佳化核心
估計、leave-one-out approximation、K-state transfer function、平均場退火理論。

This work addresses on blind separation of convolutive mixtures of independent sources. The temporally convolutive structure is assumed to be composed of multiple mixing matrices, each corresponding to a time delay, collectively transforming a segment of consecutive source signals to form multi-channel observations. As τ=1, this problem reduces to linear independent component analysis. For arbitrary τ, we propose a new algorithm to estimate the unknown convolutive structure as well as independent sources. The proposed convolutive ICA algorithm is based on optimal kernel estimation and leave-one-out approximation operated under the mean-field-annealing process. We test the new algorithm with artificially created data and two-microphone recordings of speech and musics. It is shown that the error between the estimated and given convolutive structures is significantly reduced for artificially created data and the human speech is well separated from the background musics for the two-microphone recordings. The proposed new algorithm is empirically shown effective for blind separation of convolutive mixtures of independent sources.
Keyword: Convolutive mixture, independent component analysis, blind separation, optimal kernel estimation, leave-one-out approximation, K-state transfer function, mean field annealing.

Abstract 2
1 Introduction 3
1.1 Independent component analysis . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Convolutive independent component analysis . . . . . . . . . . . . . . . . 4
2 Recurrent optimal kernel method 4
2.1 Optimal kernel estimation . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Leave-one-out approximation . . . . . . . . . . . . . . . . . . . . . . . . 8
2.3 Mean field annealing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.3.1 Two-state transfer function . . . . . . . . . . . . . . . . . . . . . 10
2.3.2 K-state transfer function . . . . . . . . . . . . . . . . . . . . . . . 10
3 Convolutive independent component analysis 12
4 Numerical simulations 13
4.1 Signal approximation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
4.2 Convolutive independent component analysis . . . . . . . . . . . . . . . . 15
4.2.1 Artificial data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
4.2.2 Blind separation of real world signals . . . . . . . . . . . . . . . . 16
5 Conclusions 16
References 17
Figure 20

[1] S. Amari, A. Cichocki, and H. H. Yang, ”A new learning algorithm for blind signal
separation”, Advances Neural Inform. Processing Syst., vol. 8, pp. 757-763, 1996.
[2] J. Basak and S. Amari, “Blind separation of a mixture of uniformly distributed
source signal: a novel approach”, Neural Computation, vol. 11, no. 2, pp. 1011-1034,
1999.
[3] A. J. Bell, and T. J. Sejnowski, “An information-maximization approach to blind
separation and blind deconvolution”, Neural Computation, vol. 7, pp. 1129-1159,
1995.
[4] J. F. Cardoso, “High-order contrasts for independent component analysis”, Neural
Computation, vol. 11, no. 1, pp. 157-192, Jan. 1999.
[5] J. F. Cardoso, "Blind Signal Separation: Statistical Principles", Proceedings of the
IEEE, vol. 86, no. 10, pp. 2009-2025, 1998.
[6] A. Hyvarinen, "Fast and robust fixed-point algorithms for independent component
analysis", IEEE Trans. on Neural Networks, vol. 10, no. 3, pp. 626-634, May. 1999.
[7] A. Hyvarinen, and E. Oja, "A fast fixed-point algorithm for independent component
analysis", Neural Computation, vol. 9, no. 7, pp. 1483-1492, Oct. 1997.
[8] D. H. Johnson, and D. E. Dudgeon, Array Signal Processing: Concepts and Tech-
niques, New Jersey: Prentice-Hall, Inc., 1993.
[9] M. Rattray, D. Saad, and S. Amari, “Natural gradient descent for on-line learning”,
Phys. Rev. Lett., vol. 81, no. 24, pp. 5461-5464, Dec. 1998.
[10] J. J. Rieta, F. Castells, C. Sanchez, V. Zarzoso, and J. Millet, "Atrial activity
extraction for atrial fibrillation analysis using blind source separation", IEEE Trans.
on Biomedical Engineering, vol. 51, no. 7, pp. 1176-1186, Jul. 2004.
[11] F. Rosenblatt, Principles of Neurodynamics: Perceptrons and the Theory of Brain
Mechanisms, Washington, DC: Spartan Books, 1962.
[12] J. M. Wu, "Natural discriminant analysis using interactive Potts models", Neural
Computation, vol. 14, no. 3, pp. 689-713, Mar. 2002.
[13] J. M. Wu and S. J. Chiu, "Independent component analysis using Potts models",
IEEE Trans. on Neural Networks, vol. 12, no. 2, pp. 202-211, Mar. 2001.
[14] J. M.Wu, and W. J. Chou, "Convolutive independent component analysis by density
estimation and leave-one-out approximation", Master thesis, Department of Applied
Mathematics, National Dong Hwa University, Jul. 2005.
[15] J. M. Wu, and K. C. Huang, "Neural optimizations by advanced mean field an-
nealing", Master thesis, Department of Applied Mathematics, National Dong Hwa
University, Jul. 2002.
[16] Z. H. Lin, "Learning generative models for density estimation and function approx-
imation", PhD. thesis, Department of Applied Mathematics, National Dong Hwa
University, Jul. 2005.
[17] J. M. Wu, M. H. Chen, and Z. H. Lin, ”Independent component analysis based on
marginal density estimation using weighted Parzen windows.”, revised for Neural
Networks, 2005.
[18] J. M. Wu, Z. H. Lin, and P. H. Hsu, "Function approximation using generalized
adalines", IEEE Trans. on Neural Networks, vol. 17, no. 3, May 2006.
[19] V. Zarzoso, and A. K. Nandi, "Noninvasive fetal electrocardiogram extraction: blind
separation versus adaptive noise cancellation", IEEE Trans. on Biomedical Engineer-
ing, vol. 48, no. 1, pp. 12-18, Jan. 2001.