Fixed-point Error Evaluation of Fast ICA and Algebraic ICA Algorithms

Independent component analysis (ICA) technique separates mixed signals blindly without any information of mixing system. The present work evaluates the error performance of fast ICA and algebraic ICA algorithms for their fixed-point implementations. Simulation study is carried on both fixed and floating point ICAs. It is observed that the word length greatly influences the separation performance. Out of the two ICAs studied the algebraic ICA offers superior performance when the same word length is used in both the cases.