Blind source separation algorithm for biomedical signal based on lie group manifold

  • Daguang Cheng School of Mechanical and Electrical Engineering, Huainan Normal University, Huainan 232038, China
  • Mingliang Zheng School of Mechanical and Electrical Engineering, Huainan Normal University, Huainan 232038, China; Human-computer collaborative robot Joint Laboratory of Anhui Province, Huainan 232038, China
Keywords: ICA; Lie group manifold; gradient descent; blind source separation
Article ID: 631

Abstract

Independent Component Analysis (ICA) is a powerful tool for solving blind source separation problem in biomedical engineering. The traditional ICA algorithm ignores the Lie group structure of constrained matrix manifold. In this paper, a gradient descent algorithm on Lie group manifold is proposed based on the geometric framework of optimization algorithm on Riemann manifold. Firstly, the orthogonal constraint separation matrices are regarded as a Lie group manifold, and the gradient of ICA objective function on the Lie group manifold is given by using Riemann metric; Secondly, the geodesic equation of the current iteration point along the gradient descent direction is calculated; Finally, a new iteration point is obtained by moving a certain step along the geodesic line, meanwhile, the step length can be adjusted adaptively. Simulation results show that the gradient algorithm on Lie group manifold is feasible for blind Source Separation, and its performance (convergence speed, stability and error) is better than other algorithms.

References

1. Yang, H. H., Amari, S. (1997). Adaptive on-line learning algorithm for blind separation-maximum entropy and minimum mutural information. Neurala computation, 7(9), 1457-1482.

2. Rinen, A. H. (1999). Surveyon independent component analysis. Neural Computing Surveys, 2, 94-128.

3. Bingham, E., Hyvärinen, A. (2000). A fast fixed-point algorithm for independent component analysis of complex valued signals. Neural System, 10(1), 1-8.

4. Ranganathan, R., Yang, T., Mikhael, W. B. (2010). Optimum block adaptive ICA for separation of real and complex signals with known source distributions in dynamic flat fading environments. Journal of Circuits, Systems and Computers, 19(2), 367-379.

5. Hyvarinen, A., Karhunen, J., Oja, E. (2001). Independent component analysis [M]. Newyork: Wiley Press, 120-124.

6. Cardoso, J. F., Laheld, B. (1996). Equivariant adaptive source separatron. IEEE Transactions on signal Processin, 45(2), 434-444.

7. Amari, S. (1998). Natural gradient works efficiently in learning. Neural Computation, 10, 251-276.

8. Amari, S. (2016). Information geometry and its applications [M]. Berlin: Springer, 10-24.

9. Hosseini, S., Pouryayevali, M. R. (2013). Nonsmooth Optimization Techniques on Riemannian Manifolds. J. Optim. Theory Appl., 158, 328-342.

10. Najmeh, H. M., Soghra, N., Mohamad, R. P. (2023). A proximal bundle algorithm for nonsmooth optimization on Riemannian manifolds. IMA Journal of Numerical Analysis, 43(1), 293-325.

11. Shen, J. Y., Huang, W. Z., Ma, J. T. (2024). An efficient and provable sequential quadratic programming method for American and swing option pricing. European Journal of Operational Research, 316(1), 19-35.

12. Wang, J. H., Wang, X. M., Li, C., & Yao, J.C. (2021). The convergence of the generalized Lanczos trust-region method for the trust-region subproblem. SIAM Journal on Optimization, 31 (1), 887-914.

13. Edelman, A., Arias, T., Smith, S. T. (1998). The geometry of algorithms with orthogonality constraints. SIAM Journal on Matrix Analysis and Applications, 20, 303-353.

14. Zhang, J. J., Cao J., Wang, Y. Y. (2013). Gradient algorithm on stiefel manifold and application in feature extraction. Journal of Radars, 2(3), 3019-313.

15. Song, R. Y. (2012). Solving sparse PCA based on the optimization methods on Riemannian manifolds [D]. Shanghai: Fudan university, 35-47.

16. Li T., Zhao J. (2017). Interference alignment based on the conjugate gradient algorithm on the Grassmannian manifold. Journal oI Yangzhou University (Natural Science Edition), 20(4), 58-62.

17. Wu, Q. F. (2012). Gradient descent on the stiefel manifold. Acta Mathematicae Applicatae Sinica, 35(4), 720-728.

18. Zhu, X. L., Zhang, X. D. (2003). Blind source separatron based on optimally selected estimating functions separatron. Journal of Xidian University, 30(3), 335-339.

19. Zhang, Q. R. (2013). Study on adaptive step size for blind source separation algorithms [D]. Xi'an: Xidian University, 9-18.

20. Liu, T. (2016). The research of blind source separation algorithm based on adaptive variable step size [D]. Taiyuan: Taiyuan University of Technology, 42-54.

Published
2024-11-19
How to Cite
Cheng, D., & Zheng, M. (2024). Blind source separation algorithm for biomedical signal based on lie group manifold. Molecular & Cellular Biomechanics, 21(3), 631. https://doi.org/10.62617/mcb631
Section
Article