Title :
Blind source separation based on variable step length natural gradient algorithm
Author :
Ji Ce ; Tang Baocheng ; Wang Jinyu
Author_Institution :
Coll. of Inf. Sci. & Eng., Northeastern Univ., Shenyang, China
Abstract :
This paper proposes an improved blind separation algorithm based on natural gradient. The traditional natural gradient algorithm is improved by a suitable nonlinear activation function and auto-adaptive step length. The effectiveness of the proposed algorithm can be shown in the simulation of speech signals separation.
Keywords :
blind source separation; gradient methods; speech processing; autoadaptive step length; blind source separation; nonlinear activation function; speech signal separation; variable step length natural gradient algorithm; Activation Function; Adaptive Step-Size; Blind Source Separation; Natural Gradient Algorithm; speech signals;
Conference_Titel :
Awareness Science and Technology (iCAST), 2011 3rd International Conference on
Conference_Location :
Dalian
Print_ISBN :
978-1-4577-0887-9
DOI :
10.1109/ICAwST.2011.6163152