MOSFC-Gradient Ascent를 이용한 엔트로피 최대화의 성능향상
Performance Improvement of Entropy Maximization using MOSFC-Gradient Ascent
3
조회
0
다운로드
1
인용
0
참고문헌
키워드
초록
본 논문은 음원분리를 수행하기 위한 방법으로 기존의 독립성분분석 방법 중 하나인 엔트로피 최대화 방법에서 빠른 수렴과 분리결과의 정확성 확보를 위해 기존 학습 방법인 Gradient Ascent알고리즘을 개선한 MOSFC-Gradient Ascent(Multiple Objects Searching and Fast Converging Gradient Ascent) 방법을 제안한다. 제안된 방법은 엔트로피가 지속적으로 증가하면 학습상수 또한 증가시킴으로써 빠른 수렴 특성을 가지며, 또한 랜덤한 서로 다른 시작 값으로 구성된 여러 개의 분리행렬을 탐색객체로 정의하고 병렬적으로 최적의 분리행렬을 찾음으로써 혼합된 신호의 수 가 증가하더라도 수렴속도의 저하 없이 높은 분리 성능을 유지하는 방법이다.
Abstract
In this paper, we propose the method of MOSFC-Gradient Ascent improving the Gradient Ascent algorithm. This methodobtains fast convergence and accuracy of separate results in Entropy maximization method being one among the existingICA(Independent Component Analysis) with the method for performing the sound source separation. If Entropy is steadilyincreased, the proposed method will have characteristic of fast convergence through increase of learning constant. Also itwill define as search object of several demixing matrix made in random other start value. As searching optimal demixingmatrixs in parallel, the proposed method is retained on high demixing performance without decrease of convergence speedeven though number of mixed signals is increased.
참고문헌 (7)
- J. M. Valin, "Robust Recognition of Simultaneous Speech by a Mobile Robot", Robotics, IEEE Transactions on, 23(4), pp.742~752 (2007)
- A. Hyvärinen, "Independent Component Analysis : Algorithms and Applications", Neural Networks, 13(4-5), pp.411~430 (2000)
- A. J. Bell, "An information-maximixation approach to blind separation and blind deconvolution", Neural Computation, 7(6), pp.1129~1159 (1995)
- J. V. Stone, "Component Analysis : A Tutorial Introduction", pp.2004
- J. P. LeBlanc, "Speech Separation by Kurtosis Maximization", Proc. ICASSP, 2, pp.1029~1032 (1998)
- S. Amari, "A New Learning Algorithm for Blind Signal Separation, In Advances in Neural Information Processing Systems", pp.757~763 (1996), MIT Press
- A. Hyvärinen, "Fast and Robust Fixed-Point Algorithms for Independent Component Analysis", IEEE Transactions on Neural Networks, 10(3), pp.626~634 (1999)